Datasets:

sulovexin

/

mysummary

like 0

journal.pntd.0007649

A high-throughput and multiplex microsphere immunoassay based on non-structural protein 1 can discriminate three flavivirus infections

Despite a marked decrease of Zika virus ( ZIKV ) infection since late 2017 , the specter of congenital Zika syndrome ( CZS ) and its re-emergence in flavivirus-endemic regions highlight the need for sensitive and specific diagnostic tests 1–4 ., Similar to the laboratory diagnosis for other flaviviruses , detection of nucleic acid as soon as possible post-symptom onset ( PSO ) is considered as the gold standard to confirm ZIKV infection , 5 , 6 ., Since many individuals test for ZIKV infection beyond the period when RNA is detectable and most ( ~80% ) of ZIKV infections are asymptomatic , serological tests remain as a key component of ZIKV confirmation 5 , 6 ., Furthermore , ZIKV can be transmitted sexually or following asymptomatic infection 7–9 ., ZIKV is a member of the genus Flavivirus of the family Flaviviridae , which includes several pathogenic mosquito-borne viruses in different serocomplexes ., The four serotypes of dengue virus ( DENV ) belong to the DENV serocomplex; West Nile virus ( WNV ) and Japanese encephalitis virus ( JEV ) to the JEV serocomplex; yellow fever virus ( YFV ) as a single member; and ZIKV10 ., Given that the envelope ( E ) protein is the major target of antibody response after flavivirus infection , different E antigens such as recombinant E protein , inactivated virions or virus-like particles have been developed for serological tests 10–13 ., Due to the presence of several highly conserved residues of flavivirus E proteins , anti-E antibodies in serum are commonly cross-reactive to different flaviviruses 13–17 ., The guidelines of Centers for Disease Control and Prevention ( CDC ) recommend that positive or equivocal results of E protein-based IgM tests require further testing with time-consuming plaque reduction neutralization tests ( PRNT ) 5 , 6 ., However , PRNT can confirm ZIKV-infected individuals who acquire ZIKV as the first flavivirus infection , known as primary ZIKV ( pZIKV ) infection , but often can only be interpreted as unspecified flavivirus infections for those who have experienced previous DENV or other flavivirus infections , limiting its application for ZIKV serodiagnosis in flavivirus-endemic regions ., When 795 sera that were IgM positive for ZIKV antigen by ELISA were tested for flavivirus neutralizing antibodies by PRNT , 45% were positive for ZIKV and at least one other flavivirus 18 ., This non-specificity may be an inherent property of the early post-infection response to ZIKV or reflect prior flavivirus experience ., A large number of Americans ( 7 million ) have experienced a WNV infection since 1999 19 and ~8 million traveled to yellow fever endemic countries in 2015 20 , 21 ., Thus , a sensitive , specific and multiplex serological test that can distinguish ZIKV and other flavivirus infections is needed in both U . S . and flavivirus-endemic countries 18 ., Moreover , several studies have shown that anti-DENV or WNV antibodies can enhance ZIKV infection in vitro 22–26 and in small animals , in which administration of DENV-immune plasma resulted in increased viremia and mortality in stat2 knock out mice 27 ., This is known as antibody-dependent enhancement , in which antibody at suboptimal concentration for neutralization can enhance DENV , ZIKV or other flavivirus entry and replication in Fcγ receptor-bearing cells such as monocytes and is believed to contribute to disease pathogenesis 28 ., Despite ADE of ZIKV by previous DENV immunity was not supported by two studies in non-human primates 29 , 30 , more in-depth studies of DENV immunity on ZIKV disease outcome and complication in humans are warranted 31–33 ., Thus , serological tests that can distinguish pZIKV infection ( p = primary ) from ZIKV infection with previous DENV ( ZIKVwprDENV , wpr = with previous ) infection are crucial to understand the pathogenesis of ZIKV and CZS in regions where ZIKV and DENV co-circulate ., Compared with traditional E protein-based assays , several enzyme-linked immunosorbent assays ( ELISAs ) based on ZIKV nonstructural protein 1 ( NS1 ) , including a recently reported blockade of binding ELISA , have shown improved specificity 34–39 ., However , secondary DENV ( sDENV ) and ZIKVwprDENV infections , of which both were common in endemic regions , cannot be discriminated 34–39 ., Moreover , none can detect and distinguish ZIKV , DENV and other flavivirus in a single assay ., With its high-throughput and multiplex ( up to 100-plex ) capacity , microsphere immunoassay ( MIA ) has been employed in the detection of cytokines , transplantation and transfusion antigens , and various bacterial and viral pathogens 40–43 ., Previously , we reported that a combination of ELISAs based on the NS1 proteins of DENV and ZIKV can distinguish various DENV and ZIKV infections 44 , 45 ., In this study , we developed a high-throughput and multiplex IgG MIA using NS1 proteins of DENV1 to DENV4 , ZIKV and WNV , and showed that the NS1 IgG MIA can detect and distinguish not only primary DENV , ZIKV and WNV infections but also sDENV and ZIKVwprDENV infections ., The Institutional Review Boards ( IRB ) of the University of Hawaii approved this study ( CHS #17568 , CHS#23786 ) ., S1 Table summarizes the numbers , serotypes , sampling time and sources of different panels of serum or plasma samples , including those from primary DENV1 ( pDENV1 ) , primary DENV2 ( pDENV2 ) , primary DENV3 ( pDENV3 ) , primary WNV ( pWNV ) , pZIKV , sDENV and ZIKVwprDENV infections as well as flavivirus-naïve individuals ., Samples collected <3 months or ≥3 months PSO were designated as convalescent- or post-convalescent-phase samples , respectively ., Samples from reverse transcription-PCR ( RT-PCR ) confirmed Zika cases were from the Pediatric Dengue Cohort Study ( PDCS ) and the Pediatric Dengue Hospital-based Study in Managua , Nicaragua between July 2016 and March 2017 46 , 47 ., The Zika cases that were DENV-naïve or previously DENV-exposed were defined as pZIKV ( p = primary ) or ZIKVwprDENV ( wpr = with previous ) panels , respectively ., The DENV-immune status was based on anti-DENV antibody testing by an inhibition ELISA at entry and annually of the PDCS 44–47 ., Parents or legal guardians of all participants provided written informed consents , and participants ≥6-year old provided assents ., These studies were approved by the IRBs of the University of California , Berkeley , and Nicaraguan Ministry of Health ., Thirty-six plasma samples from blood donors , who were tested WNV-positive by the transcription-mediated amplification ( a sensitive nucleic acid detection method used in blood bank ) , IgM and IgG antibodies between 2006 and 2015 , designated as pWNV infection , were provided by the American Red Cross at Gaithersburg , Maryland 48 ., Pre-2015-16 ZIKV epidemic convalescent- and post-convalescent-phase samples from RT-PCR confirmed cases with different primary DENV infections ( pDENV1 , pDENV2 , and pDENV3 ) or sDENV infection were from Taiwan , Hawaii and Nicaragua; 53 flavivirus-naïve samples from a seroprevalence study in Taiwan were included as control in this study 44 , 45 , 49–52 ., Samples from cases with primary DENV4 infection were not available ., Primary DENV or sDENV infection was determined by IgM/IgG ratio or focus-reduction neutralization tests as described previously 49–51 ., The NS1 gene ( corresponding to amino acid residues 1–352 ) of ZIKV ( HPF2013 strain ) with a His-tag at the C-terminus was codon-optimized ( Integrated DNA Technologies , Skokie , IL ) and cloned into pMT-Bip vector to establish a Drosophila S2-cell stable clone 44 ., ZIKV-NS1 protein from supernatants of the stable clone was purified by fast purification chromatography system ( AKTA Pure , GE Health Care Bio-Science , Pittsburg , PA ) 44 ., Purified DENV1-4 and WNV NS1 proteins were purchased from The Native Antigen ( Oxford , UK ) ., Ten μg each of the 6 purified NS1 proteins , bovine serum albumin ( BSA ) and PBS ( as negative antigen control ) were coupled individually onto 8 types of magnetic carboxylated miscrosphere beads ( 1 . 25 X 106 each ) containing different fluorophores ( MagPlexTM-C ) ( Luminex , TX , Austin ) using two-step carbodiimide process at room temperature 53 , 54 ., The antigen-conjugated microspheres were stored in 250 uL PBN buffer ( PBS with 1% BSA and 0 . 05% sodium azide , Sigma Aldrich ) at 4°C until use ., Eight types of microsphere beads coupled with different NS1 proteins , BSA or PBS were combined and diluted in PBS-1% BSA ., Fifty μL of the mixture ( containing ~1250 beads of each type ) were added to each well of a flat-bottom 96-well plate , and incubated with 50 μL diluted serum or plasma ( 1:100 dilution in PBS-1% BSA ) at 37°C for 30 min in the dark , followed by wash with 200 μL of PBS-1% BSA twice , incubation with 50 μL of red phycoerythrin-conjugated anti-human or anti-mouse IgG ( Jackson Immune Research Laboratory , West Grove , PA ) at 37°C for 45 min in the dark , and wash with 200 μl of PBS-1% BSA twice 54 ., Microspheres were then resuspended in 100 μl of PBS-1% BSA , incubated for 5 min and read by Luminex 200 machine ( Austin , TX ) ., All incubations were performed on a plate shaker at 700 rpm and all wash steps used a 96-well magnetic plate separator ( Millipore Corp . , Billerica , MA ) 54 ., Each plate includes two positive controls ( confirmed-ZIKV or DENV infection ) , four negative controls ( flavivirus-naïve samples ) , samples , and mouse anti-His mAb ( all in duplicates ) ., The median fluorescence intensity ( MFI ) was determined for 100 microspheres for each well ., The MFI values for each antigen were divided by the mean MFI value of one positive control ( MFI~104 ) and multiplied by 104 to calculate to rMFI for comparison between plates ( S1 Fig ) ., The cutoff rMFI for each antigen was defined by the mean rMFI value of 19 flavivirus-naïve samples plus 5 standard deviations , which gave a confidence level higher than 99 . 9% from 4 negatives 55 ., Each MIA was performed twice ( each in duplicate ) ., New batch of conjugated antigens was tested with flavivirus-naïve panel to determine the cutoff rMFI ., DENV1- , DENV2- , DENV3- , and ZIKV-NS1 IgG ELISAs have been described previously 44 , 45 ., Briefly , purified NS1 proteins ( 16 ng for individual NS1 protein per well ) were coated on 96-well plates at 4°C overnight , followed by blocking ( StartingBlock blocking buffer , Thermo Scientific , Waltham , MA ) , incubation with primary antibody ( serum or plasma at 1:400 dilution ) and secondary antibody ( anti-human IgG conjugated with horseradish peroxidase , Jackson Immune Research Laboratory , West Grove , PA ) , and wash 44 , 45 ., After adding tetramethylbenzidine substrate ( Thermo Scientific , Waltham , MA ) followed by stop solution , the optical density ( OD ) at 450 nm was read with a reference wavelength of 630 nm ., Each ELISA plate included two positive controls ( confirmed-ZIKV or DENV infection ) , four negative controls ( flavivirus-naïve sample ) , and samples ( all in duplicate ) ., The OD values were divided by the mean OD value of one positive control ( OD close to 1 ) in the same plate to calculate the relative OD ( rOD ) values for comparison between plates 44 , 45 ., The cutoff rOD was defined by the mean rOD value of negatives plus 12 standard deviations , which gave a confidence level of 99 . 9% from 4 negatives 55 ., Each ELISA was performed twice ( each in duplicate ) ., Two-tailed Mann-Whitney test was used to determine the P values between two groups , the two-tailed Spearman correlation test the relationship between the rOD and rMFI values , and the receiver-operating characteristics ( ROC ) analysis the cutoffs of the rMFI and rOD ratios ( GraphPad Prism 6 ) ., The 95% confidence interval ( CI ) was calculated by Excel ., We first employed the multiplex NS1 IgG MIA to test samples from primary DENV ( pDENV1 , pDENV2 and pDENV3 ) , pZIKV and pWNV infection panels ., Compared with flavivirus-naïve panel , the pDENV1 panel recognized the NS1 proteins of DENV1 ( 100% ) and other DENV serotypes ( 33 . 3 to 61 . 9% ) , but not those of different serocomplexes ( ZIKV and WNV NS1 proteins ) ( Fig 1A and 1B ) ., Similarly , the pDENV2 and pDENV3 panels recognized the NS1 protein of the homologous serotype ( DENV2 , DENV3 ) better than those of other serotypes ( Fig 1C and 1D ) , but did not recognize ZIKV or WNV NS1 protein except two samples ( recognizing WNV , 2/13 ) ., The pZIKV panel recognized ZIKV NS1 protein but not those of WNV and DENV except two sample recognizing DENV2 ( 2/38 ) , whereas the pWNV panel recognized WNV proteins rather than those of ZIKV and DENV except one sample ( recognizing DENV4 , 1/36 ) ( Fig 1E and 1F ) ., Taken together , these findings suggested that primary infection panels recognized the homologous ( infecting serotype ) NS1 protein better than other NS proteins within the same serocomplex , and in general did not recognize an NS protein of different serocomplexes ( Fig 1G ) ., We next tested samples from sDENV and ZIKVwprDENV panels ., For convalescent-phase samples , sDENV panel not only recognized NS1 proteins of DENV1-4 ( 66 . 7 to 100% ) but also those of ZIKV and WNV ( 45 . 8 to 54 . 2% ) ( Fig 2A ) ., The ZIKVwprDENV panel recognized ZIKV NS1 protein ( 100% ) as well as DENV1-4 and WNV NS1 proteins ( 60 . 0 to 90 . 0% ) ( Fig 2B ) ., A similar trend was observed for post-convalescent-phase samples ( Fig 2C and 2D ) ., These findings were in agreement with our previous reports based on NS1 IgG ELISAs 44 , 45 , and suggested that after repeated flavivirus infections , such as sDENV and ZIKVwprDENV infections , anti-NS1 antibodies cross-reacted to multiple NS1 proteins , including those from prior exposure or sometimes those with no prior exposure ., Previously we reported that sDENV panel not only recognized DENV1 NS1 protein but also ZIKV NS1 protein in IgG ELISA ( 95 . 8 and 66 . 7% , respectively ) ; similarly the ZIKVwprDENV panel recognized both ZIKV and DENV1 NS1 proteins ( 95 . 0 and 85 . 0% , respectively ) 44 ., Using the rOD ratio of ZIKV NS1 to DENV1 NS1 with a cutoff at 0 . 24 , we can distinguish ZIKVwprDENV and sDENV panels ., Since the same sDENV and ZIKVwprDENV panels recognized both DENV1 and ZIKV NS1 proteins in IgG MIA ( Fig 2A and 2B ) , we calculated the ratio of relative median fluorescence intensity ( rMFI ) of ZIKV NS1 to that of DENV1 NS1 and found that a cutoff of the rMFI ratio at 0 . 62 , as determined by ROC analysis , can distinguish these two panels with a sensitivity of 88 . 9% and specificity of 91 . 7% ( Fig 2E ) ., Since both panels also recognized DENV2 NS1 protein , we further calculated the ratio of rMFI of ZIKV NS1 to DENV2 NS1; interestingly a cutoff of the rMFI ratio at 0 . 62 was able to distinguish these two panels with a sensitivity of 94 . 4% and specificity of 90 . 9% ( Fig 2F ) ., Similar observations were found for post-convalescent-phase sDENV and ZIKVwprDENV panels; these two panels can be distinguished by a cutoff ( 0 . 62 ) of the rMFI ratio for ZIKV NS1 to DENV1 NS1 or DENV2 NS1 with a sensitivity/specificity of 90 . 0/100% or 83 . 3/100% , respectively ( Fig 2G and 2H ) ., Since these panels have been tested with individual DENV1 to DENV4 and ZIKV NS1 IgG ELISAs previously 45 , we compared the detection rates for each NS1 protein between ELISA and MIA ., For the pZIKV panel , ZIKV NS1 ELISA had a detection rate of 100% , comparable to that of MIA , for the post-convalescent-phase samples , but only 5% for the convalescent-phase samples , which was much lower than that of MIA ( 100% ) ( Fig 3A and 3B ) ., Although 19 convalescent-phase pZIKV samples were tested negative by ZIKV NS1 IgG ELISA , the relative optical density ( rOD ) values were positively correlated with the rMFI values ( correlation coefficient r = 7464 , P = 0 . 0002 ) ( Fig 3C ) , suggesting that ZIKV NS1 MIA was more sensitive than ELISA ., A positive correlation was also found between rOD and rMFI values for the post-convalescent-phase samples ( r = 8922 , P<0 . 0001 ) ( Fig 3D ) ., For pDENV1 panel , DENV1 NS1 ELISA and MIA had comparable detection rates ( 100% ) for both convalescent and post-convalescent-phase samples ( Fig 3E and 3F ) ., Similarly , a positive correlation was found between rOD and rMFI values ( Fig 3G and 3H ) ., For ZIKVwprDENV panels , ZIKV NS1 IgG ELISA and MIA had comparable detection rates for both convalescent and post-convalescent-phase samples ( Fig 4A and 4B ) ., A positive correlation was found between rOD and rMFI values for ZIKV NS1 as well as DENV1 , DENV2 , DENV3 and DENV4 NS1 tested ( Fig 4C–4E ) ., Similar observations were found for sDENV panels ( S2 Fig ) ., Table 1 summarizes the results of all samples tested with different NS1 proteins ( DENV1 , DENV2 , DENV3 , DENV4 , DENV1 , 2 , 3 or 4 , ZIKV and WNV ) in the IgG MIA ., For statistical analysis comparing different panels , one sample from each participant was included ( S2 Table ) ., The overall sensitivity of each DENV ( DENV1 , DENV2 , DENV3 ) NS1 IgG MIA to detect different DENV infections ranged from 73 . 6 to 90 . 1% and specificity from 98 . 1 to 100% ( Table 2 ) ., Interestingly , combination of four DENV NS1 IgG MIA increased the sensitivity to 94 . 5% , while maintaining the specificity of 97 . 2% , suggesting that this multiplex assay can be applied to detect DENV infections rather than distinguish different DENV serotypes ., For the ZIKV NS1 IgG MIA , the overall sensitivity was 100% and specificity 87 . 9% ., For the WNV NS1 IgG MIA , the overall sensitivity was 86 . 1% and specificity 78 . 4% ( Table 2 ) ., In this study , we developed a high-throughput and multiplex IgG MIA using NS1 proteins of DENV1 to DENV4 , ZIKV and WNV to detect and distinguish various DENV , ZIKV and WNV infections ., Based on the results , we propose an algorithm to discriminate primary DENV , pZIKV and pWNV infections , sDENV infection and ZIKVwprDENV infection ( Fig 5 ) ., Previous studies of flavivirus serodiagnosis mainly focused on two flaviviruses ., Compared with a recent study of IgG MIA containing ZIKV and DENV antigens , our multiplex IgG MIA consists of 6 antigens ( DENV1 to DENV4 , WNV and ZIKV NS1 proteins ) plus two controls ( BSA and PBS ) 56 ., To our knowledge , this is the first report of a single serological test to detect three flavivirus infections ., Our findings that combination of DENV1 to DENV4 NS1 IgG MIA increased the sensitivity to 94 . 3% while maintaining a specificity of 97 . 2% and that the rMFI ratio of ZIKV NS1 to DENV1 or DENV2 NS1 can distinguish ZIKVwprDENV and sDENV infections with a sensitivity of 83 . 3–94 . 4% and specificity of 90 . 9–100 . 0% have important applications to serodiagnosis and serosurveillance of DENV and ZIKV infections in regions where both viruses co-circulate ., Generally in agreement with our recent study of individual DENV NS1 ELISAs 45 , we found that DENV1 and DENV3 NS1 IgG MIAs can detect primary DENV infection of the homologous serotype with a sensitivity ( 100% ) higher than that for heterologous serotypes ( 25 . 0 to 100% ) ( Table 2 ) ., DENV1 , DENV2 and DENV3 NS1 IgG MIAs can detect secondary DENV infection with a sensitivity of 95 . 5 to 100% ., This was also consistent with our previous study using Western blot analysis , in which anti-NS1 antibodies recognized NS1 protein predominantly of the infecting serotype after primary DENV infection and multiple NS1 proteins after secondary infection 13 ., Taken together , due to the variable and extensive cross-reactivity of anti-NS1 antibodies after primary and secondary DENV infections , respectively , it is difficult to use a single NS1 IgG MIA or ELISA to identify the infecting DENV serotype ., Notably , the combination of four DENV NS1 IgG MIA can detect different primary and secondary DENV infections with a sensitivity of 94 . 3% and specificity of 97 . 2% ( Table 2 ) , suggesting the feasibility and application of this multiplex NS1 IgG MIA to detect DENV infection rather than distinguish DENV serotypes ., The overall sensitivity of the ZIKV NS1 IgG MIA was 100% and the specificity was 87 . 9% , primarily due to the cross-reactivity of the sDENV panel ( Table 2 ) ., The sensitivity ( 100% ) was higher than or comparable with those previously reported ( 79 to 100% ) using the Euroimmun ZIKV NS1 IgG ELISA kit 34–37 ., The ZIKV NS1 blockade of binding ELISA had an overall specificity of 91 . 4–92 . 6% , which reduced to 77 . 6–90 . 5% when comparing with sDENV panel 38 , 39 ., A recently reported ZIKV NS1 IgG3 ELISA had a sensitivity of 97% based on samples from Salvador , but it reduced to 83% when comparing with samples outside of Salvador 32 ., A previous study of multiplex IgG MIA including ZIKV NS1 reported a sensitivity of 100% and specificity of 78% for pZIKV panel based on PRNT results , however , the sDENV and ZIKVwprDENV panels were not distinguished 56 ., For the WNV NS1 IgG MIA , the overall sensitivity was 86 . 1% probably due to sampling during the early convalescent-phase for this pWNV panel ( S1 Table ) , and the specificity was 78 . 4% , mainly due to the cross-reactivity from the sDENV and ZIKVwprDENV panels ( Table 2 ) ., Using the rMFI ratio of ZIKV NS1 to DENV1 or DENV2 NS1 , we can distinguish ZIKVwprDENV and sDENV panels with a sensitivity of 83 . 3–94 . 4% and specificity of 90 . 9–100 . 0% ., This was consistent with our previous reports of IgG ELISAs using the rOD ratio of ZIKV NS1 to DENV1 NS1 or mixed DENV1-4 NS1 to distinguish these two panels with a sensitivity of 91 . 7–94 . 1% and specificity of 87 . 0–95 . 0% 44 , 45 ., It is worth noting since DENV3 and DNV4 NS1 proteins were not recognized by several samples from the sDENV and ZIKVwprDENV panels ( Fig 2A and 2D ) , they were not included in the analysis of the rMFI ratio ., Comparing the results of individual NS1 IgG MIA in this study and those of NS1 IgG ELISA reported previously 45 , we found comparable detection rates between MIA and ELISA , and positive correlations between the rMFI and rOD values for both convalescent-phase and post-convalescent-phase samples of most panels tested including pDENV1 , sDENV and ZIKVwprDENV panels except pZIKV panel ( Figs 3 and 4 and S2 Fig ) ., Of note , the IgG MIA detection rates for DENV1-4 for the post-convalescent-phase ZIKVwprDENV panel were much lower than those for the sDENV panel ( Fig 4E and S2E Fig ) , suggesting that prior DENV exposure of the ZIKVwprDENV panel may have been only to a single DENV serotype ., For the convalescent-phase pZIKV panel , the higher detection rate of ZIKV NS1 IgG MIA ( 100% ) than that of ELISA ( 5% ) and the positive correlation between rOD and rMFI values suggest that MIA was more sensitive than ELISA ( Fig 3A–3C ) ., Thus , we did not observe a trend of increased detection rates of NS1 IgG MIA from convalescent to post-convalescent phases for primary infection panels ( pZIKV , pDENV1 ) ( Fig 3B and 3F ) as previously reported for NS1 IgG ELISA and blockade of binding of NS1 ELISA 38 , 45 ., Notably we incubated 16 ng antigen coated on each well with 50 μL of serum ( 1:400 ) in ELISA , whereas we incubated ~10 ng antigen ( in 1250 beads ) with serum ( final dilution 1:200 ) per well in MIA ., The higher concentration of serum and more surface area of antigen coupled on beads may account for the higher sensitivity of the IgG MIA compared with IgG ELISA for the pZIKV convalescent-phase panel ., Although neutralization tests are still considered a confirmatory assay , they are time-consuming and can be performed only in reference laboratories ., Compared with PRNT and ELISA , the multiplex MIA requires less time ( 2 . 5 h vs . 7 h for ELISA and 5–6 days for PRNT ) and less sample volume ( 1 μL vs . 8 μL for ELISA and 144 μL for PRNT for 8 antigens or viruses ) ., The newly developed multiplex NS1 IgG MIA could have wide-ranging applications , such as serodiagnosis , blood screening , serosurveillance of ZIKV , DENV and WNV infections , and retrospective study of ZIKV infection among pregnant women with CZS 57 , 58 ., The current octaplex ( 6 NS1 antigens plus PBS and BSA controls ) IgG MIA serves as a “proof-of-concept” assay to demonstrate that NS1-based MIA can distinguish three flavivirus infections; incorporation of other antigens would increase the detection capacity for different clinical settings and studies ., These together would further our understanding of the epidemiology , pathogenesis and complications of ZIKV in regions where multiple flaviviruses co-circulate 1–4 ., There are several limitations of this study ., First , due to limited samples of < 3 months PSO from patients with primary DENV infection ( S1 Table ) , the study focused on NS1 IgG MIA ., Future studies on NS1-based IgM MIA are warranted ., Second , despite the availability of two-time point samples for the pZIKV and ZIKVwprDENV panels , future studies involving more sequential samples are needed to validate these observations ., Additionally , the sample size in each panel with well-documented infection is small ., Third , although this multiplex assay can distinguish various panels of samples with three flavivirus infections , future tests that can distinguish other pathogenic flaviviruses such as JEV , YFV and tick-borne encephalitis virus ( TBEV ) remain to be exploited 59 , 60 ., Moreover , samples with well-documented repeated flavivirus infections such as DENV with previous ZIKV infection and sequential DENV and WNV infections are lacking and remain to be investigated in the future ., In light of the successful implementation of several flavivirus vaccines and vaccine trials in flavivirus-endemic regions , serological tests that can distinguish ZIKV infection from vaccinations with DENV , JEV , YFV and TBEV vaccines are warranted 59 , 60 .

Introduction, Methods, Results, Discussion

The explosive spread of Zika virus ( ZIKV ) and associated complications in flavivirus-endemic regions underscore the need for sensitive and specific serodiagnostic tests to distinguish ZIKV , dengue virus ( DENV ) and other flavivirus infections ., Compared with traditional envelope protein-based assays , several nonstructural protein 1 ( NS1 ) -based assays showed improved specificity , however , none can detect and discriminate three flaviviruses in a single assay ., Moreover , secondary DENV infection and ZIKV infection with previous DENV infection , both common in endemic regions , cannot be discriminated ., In this study , we developed a high-throughput and multiplex IgG microsphere immunoassay ( MIA ) using the NS1 proteins of DENV1-DENV4 , ZIKV and West Nile virus ( WNV ) to test samples from reverse-transcription-polymerase-chain reaction-confirmed cases , including primary DENV1 , DENV2 , DENV3 , WNV and ZIKV infections , secondary DENV infection , and ZIKV infection with previous DENV infection ., Combination of four DENV NS1 IgG MIAs revealed a sensitivity of 94 . 3% and specificity of 97 . 2% to detect DENV infection ., The ZIKV and WNV NS1 IgG MIAs had a sensitivity/specificity of 100%/87 . 9% and 86 . 1%/78 . 4% , respectively ., A positive correlation was found between the readouts of enzyme-linked immunosorbent assay and MIA for different NS1 tested ., Based on the ratio of relative median fluorescence intensity of ZIKV NS1 to DENV1 NS1 , the IgG MIA can distinguish ZIKV infection with previous DENV infection and secondary DENV infection with a sensitivity of 88 . 9–90 . 0% and specificity of 91 . 7–100 . 0% ., The multiplex and high-throughput assay could be applied to serodiagnosis and serosurveillance of DENV , ZIKV and WNV infections in endemic regions .

Although there was a decrease of Zika virus ( ZIKV ) infection since late 2017 , the specter of congenital Zika syndrome and its re-emergence in flavivirus-endemic regions emphasize the need for sensitive and specific serological tests to distinguish ZIKV , dengue virus ( DENV ) and other flaviviruses ., Compared with traditional tests based on envelope protein , several nonstructural protein 1 ( NS1 ) -based assays had improved specificity , however , none can discriminate three flaviviruses in a single assay ., Moreover , secondary DENV infection and ZIKV infection with previous DENV infection , both common in endemic regions , cannot be distinguished ., Herein we developed a high-throughput and multiplex IgG microsphere immunoassay using the NS1 proteins of four DENV serotypes , ZIKV and West Nile virus to test samples from laboratory-confirmed cases with different primary and secondary flavivirus infections ., Combination of four DENV NS1 assays revealed a sensitivity of 94 . 3% and specificity of 97 . 2% ., The ZIKV and WNV NS1 assays had a sensitivity/specificity of 100%/87 . 9% and 86 . 1%/78 . 4% , respectively ., Based on the signal ratio of ZIKV NS1 to DENV1 NS1 , the assay can distinguish ZIKV infection with previous DENV infection and secondary DENV infection with a sensitivity of 88 . 9–90 . 0% and specificity of 91 . 7–100 . 0% ., This has applications to serodiagnosis and serosurveillance in endemic regions .

dengue virus, medicine and health sciences, enzyme-linked immunoassays, immune physiology, pathology and laboratory medicine, pathogens, immunology, microbiology, viruses, rna viruses, antibodies, immunologic techniques, research and analysis methods, immune system proteins, serology, proteins, medical microbiology, microbial pathogens, immunoassays, biochemistry, west nile virus, flaviviruses, viral pathogens, physiology, biology and life sciences, organisms, zika virus

journal.pntd.0004831

The Burden of Zoonoses in Kyrgyzstan: A Systematic Review

Zoonoses are diseases in humans , which are naturally transmissible directly or indirectly from vertebrate animals ., Of 1415 species of infectious organisms know to be human pathogens , 61% are zoonotic 1 ., The Food and Agriculture Organization of the United Nations ( FAO ) , the World Health Organization ( WHO ) and the World Organisation for Animal Health ( OIE ) have underlined the important socioeconomic impact of these diseases , yet in many low income countries the burden of zoonoses remains unknown 2 ., The lack of information often results in a vicious circle of underestimation and limited incentive to quantify the problem 3 , 4 ., Kyrgyzstan is a country in Central Asia , neighboured by China in the west , Kazakhstan in the north , and Tajikistan and Uzbekistan in the southeast ( Fig 1 ) ., Because of a poor functioning veterinary and sanitation system , emerging zoonoses are an increasing problem 5 ., Since independence in 1991 , veterinary services deteriorated , causing an increase in zoonotic disease ( ZD ) 6 , 7 ., At particular risk are the 64% of the inhabitants who live in rural areas , where livestock farming plays an important role ., Seventy-six percent of these rural dwellers are considered to be poor 8 ., The small-scale farming , which is often nomadic , allows intensive contact between humans and animals 9 ., Furthermore , in Kyrgyzstan , an estimated 100 DALYs per 100 , 000 were lost due to inadequate hygiene in 2012 10 , which ranks the country in the 61th place out of 146 low/middle-income countries on disease burden due to poor hygiene ., The combination of poor healthcare , poverty , inadequate hygiene , and the close interaction between humans , livestock and other animals , leaves a large share of the population at risk of being infected with zoonoses ., Another difficulty in assessing the burden of the diseases , is the low scientific output from Kyrgyzstan which is often published in Russian 11 , 12 ., The World Bank and the OIE have advised Kyrgyzstan to develop national animal disease control strategies 2 ., A quantification of the impact of zoonoses helps prioritizing these diseases ., The aim of this study was to quantify the burden of ZD in Kyrgyzstan using disability-adjusted life years ( DALYs ) a standardized approach to increase comparability of disease impact 13–19 ., In this review , we have assessed the available data on zoonoses in Kyrgyzstan with special attention to the potential underreporting using stochastic disease modelling ., We have comprehensively summarised the burden of the most important zoonoses that are endemic in Kyrgyzstan and addressed the underestimation in officially reported cases ., The ZDs described in this systematic review ( Table 1 ) are regarded as the most important in terms of socioeconomic impact based on the WHO report on neglected tropical diseases , the World Bank report on Kyrgyzstan of 2011 and other systematic reviews of neighbouring or overlapping regions 2 , 5 , 20 , 21 ., We assembled all the available evidence regarding prevalence or incidence of the selected ZDs in Kyrgyzstan since the country became independent ., Therefore , the time period for the search was January 1991-January 2016 ., Both formal , peer-reviewed scientific literature , and informal sources , grey literature , were considered ., A full list of sources can be found in S1 Supporting Information ., We conducted a systematic review by following guidelines of the Preferred Reporting Items for Systematic reviews and Meta-Analyses ( PRISMA guidelines , Moher , Liberati , Tetzlaff , Altman , & The PRISMA Group , 2009 ) 31 ( S2 Supporting Information– Prisma checklist ) ., A list of synonyms for the ZDs was constructed using the pathogen’s name and alternative ( popular ) names of the disease ., The computer search was constructed by combining these terms with the Boolean OR and the term ‘Kyrgyzstan’ with the AND Boolean ., The databases of PubMed , Google Scholar , Web of Science , OVID , Scopus , the WHO Global Health Library , Food and Agriculture Organization of the United Nations ( FAO ) and ProMED-mail were searched using English search terms and Google Scholar using Russian search terms ., For each database , the search construct was adapted to the specific modus operandi of the search engine ., S1 Supporting Information lists the search terms as well as the search constructs for the different databases ., Furthermore , we searched the internet for published reports on demographic surveillance sites in the English and the Russian language ., Data from government sources was contributed by the co-authors ., Following retrieval , studies were selected by critically appraising the titles and abstracts ., A study was excluded when it did not address prevalence or incidence for the specific disease , when it was not from the defined period , when it did not address the disease in humans or when it did not address Kyrgyzstan ., Secondly , the full text was screened and for each retrieved result the list of references was inspected for additional sources ( backward searching ) ., Forward searching was performed by entering the titles in google scholar using the ‘cited by’ function ., Searches were executed until no new results were found ., Additional results were screened according to the same methodology ., Finally , the selected studies were summarized based on study design , study area , disease measure ( prevalence/incidence ) and the reported margin of error ( S3 Supporting Information ) ., Each study was critically appraised on methodology , selectivity in reporting and assumptions made by the authors ., Fig 2 displays a flow diagram of the used selection strategy ., The Disability-Adjusted Life Year ( DALY ) was used as the burden-of-disease metric ., It is a health gap measure which quantifies health loss ., DALY calculation is a standardized method developed by the World Bank , Harvard School of Public Health and the World Health Organization for the Global Burden of Disease and injury ( GBD ) study and the global burden of foodborne diseases 13 , 32–35 ., It allows the comparison of health conditions across countries and across diseases 13 , 32 ., In this study , an incidence-based DALY calculation was applied ., This allows us to include all sequelae resulting from infection 36 , 37 ., The DALYs are calculated as the sum of the healthy years lost to disability ( YLD ) and the years of life lost due to premature death ( YLL ) ., The YLD is the sum of the different outcomes that result in disability , where an outcome is defined as sequelae of the disease or another categorisation of the disease , e . g . chronic vs . acute ., The YLD per outcome is the product of the duration , the incidence , and the disability weight of the outcome ., The YLL is the residual life expectancy at the age of death ., YLLs were calculated based on the life table from 38 ., YLDs with a lifelong duration were calculated based on a local life table from the WHO for Kyrgyzstan for 2013 39 ., Based on the recommendations and methodology of the GBD 2010 and FERG 33 , 36 , we have used a non-discounted and non-weighted approach in calculating DALYs ., If a disease and its outcomes were quantifiable , a corresponding disease outcome model was constructed based on literature ., When data on the incidence , the outcome of a disease , or other parameters for the DALY calculation were missing , these gaps were filled using data from neighbouring countries or overlapping regions ., The disease model or outcome-tree model was constructed per ZD using health outcomes with an evidence-based causal relationship between infection and outcome ., Disagreement in inputs of the disease model between different sources was modelled using distributions ( pert , triangular , and uniform ) accounting for this uncertainty ., A full description of the disease models , the input parameters and its uncertainties used to calculate the DALYs can be found in S3 and S4 Supporting Information ., Where available , disability weights from the GBD 2010 study were used ., Age distributions of outcomes were , when possible , based on data collected from Kyrgyzstan ., Furthermore , the total population size , age and sex distribution was obtained from census data by the National Statistical Committee of the Kyrgyz Republic and the United Nations Demographic Yearbook 40 , 41 ., The uncertainty in the estimates was modelled using Monte Carlo analysis ., We generated in this simulation 10 , 000 draws from the probability distributions ., All analyses were performed using R version 3 . 2 . 2 ( R Foundation for Statistical Computing , Vienna , Austria ) 42 ., Additional information on the analyses and disease models is provided in S4 Supporting Information ., The burden of disease was calculated for the reference year 2013 , a result of a trade-off between data availability and as recent as possible ., The data collected by the Department of State Sanitary and Epidemiological Supervision of the Ministry of Health of the Kyrgyz Republic provided the officially reported cases for notifiable diseases which includes a number of zoonoses ., This is available online 41 , 43 ., We have used the data retrieved from literature to evaluate these reported figures and assess the level of underestimation ., As reported in S3 Supporting Information , the availability of published disease data is scarce in Kyrgyzstan; we were not able to assess the burden of anthrax since not enough was known about the outcome of the cases ., In 2013 , 16 human cases of anthrax were reported ., The majority of the cases were the cutaneous form , Zoldoshev reviewed 217 cases of cutaneous anthrax with no fatalities 44 ., For alveolar echinococcosis ( AE ) , cystic echinococcosis ( CE ) , brucellosis and rabies incidence data for Kyrgyzstan was available from the Department of State Sanitary and Epidemiological Supervision of the Ministry of Health of the Kyrgyz Republic 41 , 43 ., Prevalence data on AE , CE and brucellosis was used to address the underestimation in the official data; to address the potential underestimation in rabies we have used data from overlapping regions ( Eurasia ) 27 ., The incidence of congenital toxoplasmosis was not formally recorded , but Minbaeva et al . ( 2013 ) provided estimates for Kyrgyzstan 45 ., For campylobacteriosis and non-typhoidal salmonellosis , no specific disease prevalence or incidence data was recorded for Kyrgyzstan ., The incidence estimates are based on the number of acute gastrointestinal infections reported in 2013 41 and the assumed etiological fraction as described by 21 , 22 ., This conservative estimate was used as the lower limit for the number of cases; the upper limit was formed by the etiologic proportion of the diarrhoea incidence multiplied by the gastroenteritis incidence from the European region based on Walker et al . and Lanata et al . 24–27 ., Invasive non-typhoidal salmonellosis ( iNTS ) forms an important outcome of non-typhoidal salmonellosis infection since mortality is much higher compared to the gastro-enteric manifestation of the disease 35 , 46 ., However limited data is available on the true incidence of iNTS since only few population-based incidence studies have been conducted ., Therefore , we have used the ratio between iNTS:NTS as described by Ao et al . 46 who classified Kyrgyzstan in the Asia/Oceania region where the proportion iNTS:NTS was 1:3 , 851 compared to 1:7 in European region which included Russia; the global average ratio was 1:28 46 ., We identified 438 unique citations and excluded 411 by title and abstract screening ., Of the remaining 28 potential eligible citations with relevant abstracts , 10 were eligible for full text review ., The PRISMA flowchart summarizing the data collection process is presented in Fig 2 ., Reports published during January 1991-January 2016 were searched ., The last search was performed on 19-02-2016 ., All collected data are summarised in S3 Supporting Information ., In 2013 seven ZDs were quantifiable in Kyrgyzstan ., AE , brucellosis , campylobacteriosis , CE , congenital toxoplasmosis , NTS and rabies ., These were responsible for an estimated total of 141 , 583 33 , 912–250 , 924 new cases resulting in 35 , 209 13 , 413–83 , 777 DALYs and 576 279–1 , 168 deaths ( Table 2 , Fig 3 ) ., Both Rabies and AE contribute a large number of DALYs per case , 70 . 1 10 . 0–90 . 0 DALYs/case and 50 . 3 20 . 7–78 . 3 DALYs/case respectively , due to high mortality ( Fig 4 ) ., Campylobacteriosis and NTS had relatively low mortality but a high incidence; most of the mortality was due to the sequelae Guillain Barre Syndrome ( GBS ) and invasive non-typhoidal salmonellosis ( iNTS ) respectively , see S4 Supporting Information ., Infections with salmonellosis and AE were responsible for the majority of deaths , respectively 254 66–571 and 236 153–466 ., Although only 5 . 1% ( 11/216 ) of the cases of congenital toxoplasmosis was fatal , the DALY/case is high ( 6 . 69 ) due to early onset of the sequelae and the lifelong duration ., Table 2 provides the estimates for the number of cases , the DALY , the number of deaths and the DALY per case per disease and their 95% uncertainty range ., Fig 3 displays a graphical representation of the per annum burden per disease and its uncertainty , plotting the DALY estimates from Table 2 per disease ., Table 3 displays the sensitivity analysis with different iNTS:NTS ratios ., Fig 4 provides the percentage of YLD and YLL per disease ., Premature mortality , or YLL , ( in blue in Fig 4 ) contributes for all diseases most to the DALY , ranging from 67% for CE to 100% for Rabies ., This work provides a first attempt at quantifying the burden of ZD in Kyrgyzstan ., It underlines the lack of published data on many zoonoses in this region ., However , the estimates of the impact of the ZDs help to break the vicious circle of underreporting by providing estimates of the true incidence and burden of these diseases ., Because of the scarcity of data we did not exclude information based on methodology; we analysed it using conservative assumptions and stochastic modelling to handle uncertainty 47 ., We have used official Kyrgyz data and addressed its underestimation ., The total burden of the seven quantified ZDs ( 35 , 209 13 , 413–83 , 777 DALYs in 2013 ) is slightly less than the yearly burden of HIV , which was attributable for 38 , 870 21 , 261–64 , 297 DALYs in 2010 in Kyrgyzstan 48 ., This burden is based on prevalence based DALYs , as used in the GBD 2010 studies 33 ., Forty-three percent of the estimated burden of zoonoses or 14 , 967 6 , 213–32 , 319 DALYs in 2013 , in Kyrgyzstan is caused by echinococcosis ( both AE and CE ) ., Torgerson et al . estimated in 2010 that in China 16 , 629 new cases of AE per year arose among 22 . 6 million people at risk 17 ., In Kyrgyzstan we estimate for 2013 that 236 cases arose among 5 . 7 million people ., However AE is characterized by a clustered distribution and some regions have a much higher incidence rate 49 ., The officially reported incidence of AE has increased since 2004 at an alarming rate ., Where before 2004 only 0–3 cases per year were reported , in 2013 148 cases were officially reported 49 , 50 ., We assume that , corrected for underestimation , the incidence is likely to be approximately 236 153–466 cases in 2013 ., Although the goal of this study was to provide an estimate of the burden of zoonoses in 2013 , the collected data allows us to reflect on temporal trends ., Raimkylov & Kuttubaev , and Usubalieva et al . describe an increasing trend over the last decade in the incidence of echinococcosis 49 , 50 , although increased awareness might lead to more diagnoses ., The yearly incidence of brucellosis , on the other hand , seems to decline over time ( S3 Supporting Information ) ., The application of the ocular Rev-1 vaccination over several years has most likely resulted in this decreased incidence 51 ., Over time , the improvement of diagnostics and the application of novel treatments may cause changes in the outcome of the disease and thus the DALY per case ., For example , in our analysis , we assumed that all cases of AE are eventually fatal due to insufficient treatment ., However , as illustrated in Switzerland , adequate treatment of the disease will lead to an increased survival 52 ., This illustrates the need for periodic updating of the burden assessment ., We choose an incidence-based approach in the DALY modelling because it allows us to include all sequelae resulting from infection ., However , one of the consequences of using the incidence-based DALY approach , is that deaths in the future are attributed to the year of the infection ., Careful interpretation of the mortality rate is therefore advised ., For example , AE did not cause the reported number of deaths in 2013 since its incidence is increasing and it has a long latency; the deaths that will be caused by the infections diagnosed in 2013 are attributed to that year ., For a disease with a short incubation time and a relative constant incidence rate , such as rabies , the difference is not so striking ., Using a prevalence based DALY approach in diseases with a trend over time and a long latent phase or incubation period , might lead to under or overestimation as it reflects past infection rather than a present day event 37 ., Another limitation is the assumption , in common with other burden studies , that the outcome of diseases can be extrapolated to different countries ., Regional differences in pathogens might change the tropism of the causative agent or cause a shift towards certain sequelae ., The same holds true for other spatially fluctuating factors , such as co-infection; The incidence of iNTS , for example , is correlated with malaria and HIV infection 46 ., This underlines the importance of not only reporting incident cases , but also of documenting disease outcome ., Other factors such as ethnicity might also have an influence on disease outcome , as for example has been postulated in tuberculosis and Plasmodium falciparum malaria 53 , 54 ., Even more striking are the vast differences in treatment according to region , as illustrated for AE ., Likewise , brucellosis , with inadequate treatment is more likely to become chronic or relapse 55 which increases the burden ., Underestimation of disease is caused by under-ascertainment and underreporting of cases ., A disease might not be severe enough for the patient to visit a medical facility ., In addition , patients might have limited access to care or the disease are not accurately diagnosed ., Underreporting is the result of incomplete registration of cases ., Even in countries with a high standard of medical care , such as WHO high-income countries , reported cases form only the tip of the iceberg of the true incidence ., For example , it is estimated that only 1/30 . 3-1/86 cases of campylobacteriosis are reported in the USA 56 , 57; the estimate in the European Union is that on average 1/47 cases of campylobacteriosis are reported 58 ., CE is often substantially underreported ., In Uzbekistan the official case numbers appear reported were 1 , 435 cases reported in 2000 and 819 cases reported in 2001 59 ., However , Nazirov and others undertook a detailed study of hospital records throughout Uzbekistan and found a total of 4 , 430 cases in 2000 and 4 , 089 cases in 2001 ., Likewise in Chile official notifications between 2001 and 2009 were a mean of 311 cases per annum , whilst a detailed audit of hospital records revealed a mean of 1 , 009 cases per annum 60 ., The assumption we made on the underestimation of the incidence of ZDs is conservative ., For most ZDs we have used either a uniform or a pert distribution and included the officially reported incidence as minimum and the with a multiplication factor corrected value as maximum ., The assumption for the mode in AE , CE and brucellosis are also conservative ., The multiplication factor we used to correct for underestimation in brucellosis ( 4 . 6 ) lies close to the mean multiplication factor ( 5 . 4 1 . 6–15 . 4 ) Kirk et al . used in 35 ., Hampson et al . reviewed the global burden of Rabies and estimated for 2010 and estimated 14 rabies deaths in Kyrgyzstan contributing to 887 DALYs 27 ., Since rabies is a fatal disease , which often affects the young , it is possible that some cases go unreported in Kyrgyzstan ., We believe that our estimate and its 95% uncertainty range represent the true incidence ., The burden consists only of the estimated fatal cases , and not the disability caused by dog bites and the burden of the treatment ., This indirect burden is highest in countries where crude nerve-tissue vaccines are used 61 , which is not the case in Kyrgyzstan ., Other carnivores than dogs , are assumed not be relevant in contributing to the transmission risk 62 , however , there is a steep increase described in the wolf population and an increasing contact rate between humans and these wild carnivores in mainly in the south of Kyrgyzstan 63 ., In this study we have explored the proportion of diarrhoea attributable to Campylobacter and non-typhoidal Salmonella in Kyrgyzstan ., We assumed etiologic proportion of diarrhoea of both pathogens based on literature 64 ., Close inspection of the reported incidence of acute intestinal infections , reveals an approximate two-fold increase in cases between 2004 and 2007 ., However , the change was likely because the funding of hospitals was modified to a case-based system 65 ., This illustrates that the variance in reported data does not always represent epidemiological change; it can be merely a reflection of an alteration in policy ., The estimates we present are based on conservative extrapolates from overlapping regions ., However , more accurate incidence data on salmonellosis and campylobacteriosis in Kyrgyzstan are lacking ., Estimates of overlapping regions often lacked nuance and tend to group heterogeneous countries ., Our median incidence estimates for both NTS ( 1 , 101/100 , 000 ) and campylobacteriosis ( 1 , 305/100 , 000 ) are higher than the estimated yearly incidence by Havelaar et al . of campylobacteriosis ( 802/100 , 000 cases ) and of NTS ( 318/100 , 000 cases ) in the EUR B region 66 ., However earlier estimates by the same author are higher; ranging from 1 , 800–11 , 800 cases/100 . 000 for salmonellosis and 2 , 240–13 , 500 cases/100 . 000 for campylobacteriosis 58 ., The data presented in 58 shows a correlation between Gross Domestic Product ( GDP ) and both salmonellosis and campylobacteriosis; both diseases have a higher estimated true incidence in countries with a lower GDP ., There seems to be no clear relation between the quantity of consumed protein ( egg , chicken , and pork ) according to the FAO and the estimated true incidence of the two diseases in the different European countries ( EU-27 ) 58 ., We believe that although chicken , egg and pork consumption in Kyrgyzstan are lower than in EU-27 countries , the lower GDP and the lower hygiene standard in Kyrgyzstan justify our estimates ., To obtain more reliable burden estimates of both campylobacteriosis and salmonellosis , it would be advisable to undertake a community-based incidence study in Kyrgyzstan ., Both diarrhoea incidence and aetiology are important inputs to narrow the uncertainty around our estimates ., Furthermore , a longitudinal study on the aetiology of febrile illness might provide a reliable estimate of the burden of different zoonoses or sequelae ( brucellosis , iNTS , listeriosis , Q-fever , leptospirosis ) ., A small scale investigation in Bishkek showed that part of the undiagnosed febrile illness was due to brucellosis and Q-fever 67 ., To date , the exact burden of leptospirosis in Kyrgyzstan is unknown ., Although occurrence of leptospirosis in cattle in Kyrgyzstan has been reported 68 , no data on the occurrence of this ZD in humans in Kyrgyzstan is available ., Torgerson et al . estimated that the burden of leptospirosis in Kyrgyzstan was 927 355–1629 DALYs per year 69 ., Costa et al . clearly illustrate a lack of data on the occurrence of leptospirosis in Central Asia; the estimates of incidence for Kyrgyzstan were based on extrapolation using a multivariable regression model 70 ., Likewise , the role of cryptosporidium and giardia as causative agent for ZD in Kyrgyzstan has not been established ., These parasites have zoonotic potential 71 , however the incidence of the disease caused by these parasites has not been investigated in Kyrgyzstan , nor has the role of animals in the transmission of these ZDs been quantified ., Therefore , burden assessment at this moment is not feasible ., Only sequelae that have a solid proven causal relationship with the pathogen have been included in the disease models we used ., Reactive arthritis , irritable bowel syndrome and GBS are evidence based sequelae of campylobacteriosis 72 ., However , we followed the conservative assumption of Kirk et al . that the relation between some sequelae were not sufficiently proven in middle and high-mortality countries 35 ., Most of the burden of salmonellosis is due to YLLs , mainly deaths caused by iNTS ., The sensitivity analysis ( Table 3 ) illustrates the influence of the proportion of iNTS:NTS ., This underlines the importance of investigating the incidence of iNTS in Kyrgyzstan and is in line with the findings of Ao et al . 46 conclude that there is a lack of population-based incidence data on iNTS ., In a limited-means setting such as Kyrgyzstan it is inevitable for policy makers to prioritize health care needs ., The DALY provides one tool to do so , but is by itself not sufficient 73 ., In the application of the DALY by healthcare legislators , it is important to look at the presented figures in a wider context ., DALYs should be combined with for example , economic parameters in cost-utility analyses 74 ., It is also important to realize that the DALY might not capture the full effect of the disease and that a disease might have bigger impact than just on the ones directly affected 75 ., Especially ZDs often cause economic loss in livestock production as well 21 , 76 ., Where in this paper we have only quantified the human burden , it makes sense to extend the work with the assessment of the economic impact of the disease in both humans and animals ., Furthermore , it is advised to conduct an integrated approach in disease intervention and prevention where both veterinary and human health officials work together 3 .

Introduction, Materials and Methods, Results, Discussion

Zoonotic disease ( ZD ) pose a serious threat to human health in low-income countries ., In these countries the human burden of disease is often underestimated due to insufficient monitoring because of insufficient funding ., Quantification of the impact of zoonoses helps in prioritizing healthcare needs ., Kyrgyzstan is a poor , mountainous country with 48% of the population employed in agriculture and one third of the population living below the poverty line ., We have assessed the burden of zoonoses in Kyrgyzstan by conducting a systematic review ., We have used the collected data to estimate the burden of ZDs and addressed the underestimation in officially reported disease incidence ., The estimated incidences of the ZDs were used to calculate incidence-based Disability Adjusted Life Years ( DALYs ) ., This standardized health gap measure enhances comparability between injuries and diseases ., The combined burden for alveolar echinococcosis , cystic echinococcosis , brucellosis , campylobacteriosis , congenital toxoplasmosis , non-typhoidal salmonellosis and rabies in Kyrgyzstan in 2013 was 35 , 209 DALYs 95% Uncertainty interval ( UI ) :13 , 413–83 , 777; 576 deaths 95% UI: 279–1 , 168 were attributed to these infections ., We estimate a combined median incidence of ZDs of 141 , 583 cases 95% UI: 33 , 912–250 , 924 in 2013 ., The highest burden was caused by non-typhoidal Salmonella and Echinococcus multilocularis , respectively 14 , 792 DALYs 95% UI: 3 , 966–41 , 532 and 11 , 915 DALYs 95% UI: 4 , 705–27 , 114 per year ., The health impact of zoonoses in Kyrgyzstan is substantial , comparable to that of HIV ., Community-based surveillance studies and hospital-based registration of all occurrences of zoonoses would increase the accuracy of the estimates .

Zoonoses are diseases transmitted from vertebrate animals to humans ., They can cause a variety of symptoms ranging from mild gastrointestinal complaints to debilitating illness and even death ., Especially in low-income countries where animals play an important role for many , the burden of these diseases can be substantial ., However , there is often little attention for these diseases , thus they remain under-researched and underfunded ., In this review , we present estimates of the burden of the most important zoonotic diseases in Kyrgyzstan for the reference year 2013 ., We estimated the burden by calculating the incidence-based disability adjusted life years ( DALYs ) , allowing comparison between diseases and injuries ., Disease frequency data is scarce and hospital-based incidence data often underestimates the true incidence of the disease ., By addressing the underestimation in officially reported incidence using data from our systematic review , we estimated the true incidence of the most important zoonoses in Kyrgyzstan ., We quantified the substantial impact these diseases have on the wellbeing of people in Kyrgyzstan in 2013 ., The results underline the need for more intensive monitoring and surveillance of zoonotic diseases .

medicine and health sciences, tropical diseases, parasitic diseases, salmonellosis, brucellosis, bacterial diseases, rabies, neglected tropical diseases, veterinary science, campylobacteriosis, echinococcosis, public and occupational health, infectious diseases, veterinary diseases, zoonoses, helminth infections, biology and life sciences, viral diseases

journal.pcbi.1000611

Distributed Dynamical Computation in Neural Circuits with Propagating Coherent Activity Patterns

To understand brain function , it is essential to study the collective electrical activity of neural circuits 1 ., This activity typically exhibits intriguing spatiotemporally organized patterns: they are commonly observed in multi-unit electrophysiological recording , EEG local field potential recording , MEG , optical imaging and fMRI imaging , both in spontaneous activity 2–5 and evoked responses 6–21 ., In space , these patterns often take the form of localized patches or clusters of activity 2–16 ., Recordings over large populations of neurons have shown that several of such localized patterns can occur simultaneously across cortical regions 2–16 ., Over time , these patterns often do not remain at specific locations ., As self-sustained entities , they propagate or move about in space 4–8 , 10–16 ., In doing so , they interact with each other , resulting in dynamical collective behavior ., Here we will consider what kind of functional role this behavior may have ., Propagating coherent patterns have been registered in the experimental literature as “spreading” or “drifting” activity 4–8 or as “traveling waves” 13–24 ., The simultaneous presence of several of these patterns has been observed in the spontaneous activity of cat visual cortex 4 , 5; see 25 for a corresponding model study , in evoked response patterns in turtle olfactory bulb 14 , and visual cortex of various species 9 , 15 , as well as in sensorimotor cortex of behaving mice 7 ., When several localized , moving patterns occur together , they are likely to interact ., Indeed , interactions have been shown to occur in rat somatosensory cortex 13 ., To describe the collective activity in olfactory , visual , auditory and somatosensory cortices of behaving rabbits , the term “interacting wave packets” was explicitly used 11 , 12 , which nicely captures the relevance of propagations and interactions of these patterns ., Despite the ubiquity of these patterns and their interactions , their fundamental functional role has remained unknown ., Although some authors have speculated on the role of propagating waves 26 , the functional implications of other aspects such as the simultaneous presence of multiple propagating patterns or their interactions have remained completely unclear ., Current theoretical frameworks describe neural activity either in computational or dynamical systems perspectives ., Conventional computational theory is based on the manipulation and representation of static symbols 27 ., This perspective contradicts the temporal variability of brain activity , which calls for a dynamical systems approach ., When dynamical systems theories are applied to neuroscience , the prevailing concept is that of stable low-dimensional attractors 28 ., This notion , although it has provided many important insights , is less suitable to capture the functional role of brain activity in its actual spatiotemporal complexity ., We need to resolve the restrictions of conventional computation and standard dynamical systems theories , in order to describe neural activity and understand its fundamental function ., This study is based on the consideration that neural circuits are spatially-extended , pattern-forming systems , containing large numbers of simple neurons with spatially restricted connectivity 29 , 30 , 31 ., In spatially extended physical systems composed of large numbers of simple interacting elements , such as reaction-diffusion systems and fluidic systems , localized propagating coherent patterns are a common feature known under different names , including wave packets , spots , breathers and soliton waves , amongst others 32 , 33 , 34 ., They are an emergent , collective property of these systems ., Using these systems as analogy , we construct a simple , spatially extended neural circuit model to represent the gross architecture within the cerebral cortex ., As an emergent , collective property of the system , the circuit exhibits dynamical activity patterns , reproducing some of the complexities observed in empirical studies ., In particular , the circuit provides simultaneous propagation of multiple locally coherent patterns and their interactions ., By revealing how their ongoing collective behavior can naturally embody computation , we demonstrate what fundamental function these patterns can serve ., Propagating coherent spiking patterns can support several essential aspects of a computational processing ., As self-sustained objects , these patterns can signal information by propagating across neural circuits ., Information processing , or computation , occurs when they interact or , specifically , collide with each other ., Collectively , these patterns perform distributed , parallel and cascaded computational operations , thereby enabling neural systems to work in an efficient and flexible way ., We shall call this distributed dynamical computation , which is proposed as a framework for understanding spatiotemporal propagating activity patterns in neural circuits ., This understanding links their dynamics with a form of non-conventional , abstract computation ., The importance of spatiotemporal dynamical patterns in the brain has been proposed in 29 , with an emphasis on spatial modes and their coupling ., Here , we have focused on propagating coherent activity patterns , which are ubiquitous in the brain ., These dynamical patterns are neither random nor stable; rather they are characterized by rich dynamical behaviors ., We have used a simple , stereotypical spiking neural circuit to generate spatially localized propagating patterns ., The patterns capture some of the key features of real pattern complexities: a distribution of multiple localized activity patterns , their propagations and their mutual interactions ., To understand their fundamental functional role , we propose the notion of distributed dynamical computation ., Localized propagating patterns are the underling primitives of dynamical computation; over time they transfer information across space and process information through their interactions ., Collisions distributed over different locations and occurring at different time moments can be connected to each other by propagating patterns ., This mechanism enables elementary computations to occur in a cascaded fashion , resulting in more complex computations ., In addition , several interactions distributed across different locations can occur simultaneously , resulting in parallel processing ., Dynamical computation emerges on the basis of activity in neural circuits; they enable and sustain propagating localized patterns and their interactions ., In this framework , the propagation and processing of signals are fluid; signals do not rely on the fixed physical lines of neural circuits to guide their propagation trajectories ., The computations are not confined to specific anatomical sites; rather they occur wherever moving patterns collide with each other ., With respect to real-brain architecture , this is clearly a simplification ., We may consider the neural architecture as biasing the trajectories of propagating patterns to various extents ., Nevertheless , a certain independence of fixed connectivity structure must be at the basis of how flexibility in brain functions is achieved ., Propagating coherent activity patterns implement logical operations in a manner reminiscent of the collisions in the classical billiard ball model 45 , 46 , 47 ., In this model , however , collisions are elastic and reversible , whereas in our model they are inelastic and therefore irreversible ., This allows the exchange of information between the interacting patterns ., The corresponding computations are equally irreversible and therefore context-dependent ., An additional essential difference with the billiard ball model is that computation in our model is naturally embedded in the ongoing behavior of a circuit ., Computation based on the propagations and interactions of coherent spiking patterns in neural systems is definitely a non-conventional form of computation ., Conventional computation requires information to be represented and manipulated in the form of static symbols 27 ., As the longstanding debate between computationalists and dynamicists 51 has pointed out , static symbols are less suitable to describe the temporal variability in the way the brain executes its functions and how it achieves flexibility ., Dynamical computation can capture the spatiotemporal characteristics of brain activity patterns and provide them with an underlying computational interpretation ., By synthesizing dynamics and computation , the present approach offers a starting point for a comprehensive understanding of the working mechanisms of the brain ., The collective propagation of activity patterns through a substrate of neurons can be portrayed as spatiotemporal spike chains ., Our current emphasis on propagating patterns bears a similarity to the paradigm of synfire chains 52 , 53 , in which sequential spike chains play a central role ., They are obtained by setting up feed-forward networks , designed to support wave-like spikes propagation through them ., These networks perform information processing by synchronizing different spike chains 52 , 53 ., In our model , spatiotemporal spike chains are an emergent property of recurrent networks 54 ., Rather than synchrony , their nonlinear pattern-forming capacities and transient interactions are the essential mechanisms for dynamical computation ., In the current study we have mainly focused on general-purpose computation based on ongoing , autonomous dynamics of neural circuits ., We have also found that external perturbations can modulate the ongoing patterns , which include their propagations and interactions ., Hence , propagating activity patterns could enable neural systems carry out some specific computations when actual sensory inputs are given ., Indeed , propagating coherent patterns such as propagating waves have been found in evoked activity 7 , 9 , 11 , 12 , 14 , 55 ., Furthermore , during whole computing processes based on propagating coherent patterns , internal synaptic modifications and external feedbacks from other parts of the brain can be used to shape or control dynamical wave pattern to generate specific propagating patterns as required outputs or behavior sequences ., The effect from feedback activity is analogous to use feedback signals to control waves patterns in spatially-extended non-equilibrium physical systems 56 ., Instead of focusing on multiple , stationary patch patterns 57 , 58 or single propagating wave pattern as in the most studies about neural fields 58 , 59 , the current model generates dynamical spiking activity patterns that can capture some of the complexities of empirically observed patterns ., Therefore , the current study provides specific , experimentally testable predictions ., In particular , the collective behavior of interacting , propagating coherent patterns belongs to the class of anomalous super-diffusion ., As a process with underlying long-range coherence , collective anomalous super-diffusion is an important indicator of complicated , nontrivial interactions between propagating patterns ., Its presence can be tested experimentally in a straightforward way ., First qualitative indications that this process may occur in real neural circuits are the seeming randomness of the points of origin of neural activity patterns and the variability of their propagating directions 11 , 22 , 23 ., More conclusive evidence can be obtained through calculating the MSD of the collective motions in the same way as for the current model data ., In the current dynamical computational framework , propagating coherent activity patterns are the fundamental primitives for signaling information and for processing information through their interactions ., Indeed , at the level of neural circuits , signaling information by propagating coherent patterns has been clearly and very well documented as an important component of the function of the cortex 18–21 ., Interactions between multiple active patterns , however , have merely been registered in experimental studies without considering their importance 11 , 12 , 13 ., Our current work shows in an abstract , principled way how these interactions could play a key role in dynamical computation ., For instance , in the visual cortex of ferrets , top-down influences have been found to be evident in terms of localized wave patterns 17 , which could have collisions with wave patterns evoked by external visual inputs; such collisions might reflect “attention guided” processing of visual stimuli ., It is , therefore , of crucial importance to study interactions between different propagating wave patterns experimentally and sow how they relate to the functions of the cortex .

Introduction, Discussion

Activity in neural circuits is spatiotemporally organized ., Its spatial organization consists of multiple , localized coherent patterns , or patchy clusters ., These patterns propagate across the circuits over time ., This type of collective behavior has ubiquitously been observed , both in spontaneous activity and evoked responses; its function , however , has remained unclear ., We construct a spatially extended , spiking neural circuit that generates emergent spatiotemporal activity patterns , thereby capturing some of the complexities of the patterns observed empirically ., We elucidate what kind of fundamental function these patterns can serve by showing how they process information ., As self-sustained objects , localized coherent patterns can signal information by propagating across the neural circuit ., Computational operations occur when these emergent patterns interact , or collide with each other ., The ongoing behaviors of these patterns naturally embody both distributed , parallel computation and cascaded logical operations ., Such distributed computations enable the system to work in an inherently flexible and efficient way ., Our work leads us to propose that propagating coherent activity patterns are the underlying primitives with which neural circuits carry out distributed dynamical computation .

The brain processes information with extraordinary efficiency , and can perform feats such as effortlessly recognizing objects from among thousands of possibilities within a fraction of a second ., This is accomplished because the brain represents and processes information in a distributed fashion and in a dynamical way ., This processing is manifested in spatiotemporal neural activity patterns of great complexities within the brain ., Here , we construct a spiking neural circuit that can reproduce some of the complexities , which are evident in terms of multiple wave patterns with interactions between each other ., We show that their dynamics can support propagating pattern-based computation; spiking wave patterns signal information by propagating across neural circuits , and computational operations occur when they collide with each other ., Such dynamical computation contrasts sharply with that done by static and physically fixed logic gates operating in other computing machines such as computers ., Moreover , we elucidate that the collective dynamics of multiple , interacting wave patterns enable computation processing implemented in a fundamentally distributed and parallel manner in the neural circuit .

biophysics/theory and simulation, computational biology/computational neuroscience

journal.pgen.0030147

Adaptive Evolution of Conserved Noncoding Elements in Mammals

Phenotypic evolution proceeds both by changes in protein coding sequences and by changes in gene expression that determine when , where , and how much genes are expressed 1–3 ., Although recent genome-wide studies have begun the process of identifying genes that show signals of adaptive evolution in coding sequences 4 , much less is known about the adaptation of regulatory sequences ., One avenue to studying adaptation of gene regulation is to identify regulatory elements that show rapid evolution at the DNA sequence level 2 ., However , a challenge for this approach is that at present we have only limited knowledge of the DNA sequence elements that drive gene expression and regulation ., One possible way forward is to study the evolution of conserved noncoding elements ( CNCs ) 5–7 ., In recent years it has been shown that ∼3 . 5% of noncoding DNA sequence is substantially conserved across diverse mammals 8–10 , and that a smaller amount of noncoding sequence is also shared with more distant vertebrates , including chicken and even fish 9 , 11–13 ., Some CNCs show extremely high levels of conservation; for example , Bejerano et al . 9 identified 481 segments longer than 200 bp that are absolutely conserved among the human , rat , and mouse genomes ., Recent studies of CNCs , using varied definitions , have reported that most CNCs are segments of around 100–300 bp , and that they are widely distributed across the human genome 9 , 10 , 14–18 ., CNCs are not preferentially located near genes 18 ., In some cases , clusters of CNCs are found in gene deserts and a subset of these CNCs have been shown to play functional roles as enhancers 19–21 ., It has been shown repeatedly that screening for CNCs is an effective method for identifying cis-regulatory modules of gene expression 18–25 ., CNCs that are shared among humans and distant outgroups such as Fugu are heavily overrepresented near developmental regulator genes , and many serve as highly conserved regulators of these functionally conserved genes 13 ., That said , there is still considerable uncertainty about the function of most CNCs , and it has been suggested that some CNCs may serve other kinds of functions , perhaps including roles in chromatin structure or structural connections between chromosomes 26 ., In principle , another possibility might be that many CNCs could simply be regions of the genome with low mutation rates ., However , two kinds of evidence argue convincingly that the low evolutionary rates of CNCs are indeed due to selective constraint ., First , the allele frequency spectrum of human SNPs that lie within CNCs is skewed towards rare variants , consistent with the action of weak purifying selection 27 , 28 ., Second , the rate of evolutionary change of CNCs is closer to the neutral rate in primates than in rodents 28 , 29 ., The latter observation is probably due to reduced efficiency of weak purifying selection in primates , which have smaller effective population sizes ., Hence , in this study , in view of the likely functional importance of CNCs , we set out to describe the patterns of evolutionary sequence change in these elements ., We start with a simple null model in which the evolution of each CNC is characterized by a single substitution rate parameter r that accounts for varying levels of constraint and local mutation rate across CNCs ., For each CNC we compare the null model to a hierarchy of alternative models that allow the CNC to have different evolutionary rates in different parts of the phylogeny ., In the simplest alternative model , the CNC evolves at a single rate across the phylogeny except for one branch , which shows a change in rate ( Figure 1 ) ., More complex alternative models allow multiple changes in rate ., Increases in rate can be interpreted as evidence for positive adaptation or relaxation of functional constraint for the element in question ., Decreases in rate are consistent with a tightening of selective constraint ., Two recently published papers 5 , 7 have taken similar approaches to identify nongenic regions that show accelerated evolution specifically in the human lineage ., Both studies concluded that human lineage-selection signals are enriched near neurological genes ., In the study of Pollard et al . 5 , the most dramatically accelerated region was found to be part of a novel RNA gene that is expressed during cortical development ., Here , we expand this kind of approach to look more broadly at evolutionary patterns of CNCs across the mammals ., To identify CNCs that have been targets of selection , we introduce a likelihood ratio test that we call the “Shared Rates Test” ( SRT ) ., Under the null model , the divergence times of lineages are shared across CNCs , but each CNC may evolve faster or slower according to its local mutation rate and level of evolutionary constraint ., For each CNC , we test whether any branches are surprisingly long or short compared to the others , indicating speed-ups or slow-downs of the substitution rate ., For example , in Figure 1 , the first two trees evolve at different rates , but with the same tree “shape” ( i . e . , the ratios of branch lengths are the same ) ., In contrast , the third tree has a longer-than-expected branch on the human lineage , suggesting the action of natural selection ., In our model , each branch of the mammalian tree has a branch-length parameter vb , defined as the average number of substitutions per site on branch b for CNCs evolving under a constant level of constraint ., ( Here , vb is defined as the average number of substitutions per site on branch b across all CNCs . ), In addition , under the null hypothesis , each CNC is associated with a single rate parameter r0 ( h ) ( where h indicates a particular CNC ) ., Then the number of substitutions that occur in CNC h , on branch b has an expectation at each site of Nb , h , where, Under the null model , there are seven branch length parameters for the tree that we consider , and one additional rate parameter for each CNC ., As described in the Methods and Text S1 , we obtain a joint maximum likelihood estimate for all the parameters , assuming the Felsenstein 84 model of sequence evolution 31 ., Our model is designed so that all CNCs have the same expected tree shape ( i . e . , the ratios of expected branch lengths are the same ) ., However the total size of the tree is allowed to vary according to r0 ( h ) , in order to reflect variation in mutation rates and the level of selective constraint across CNCs ., In addition , we place no constraints on the relative values of the vb , so that lineage-specific variation in mutation rates ( such as the higher substitution rate in rodents ) is reflected in longer estimates for those branch lengths ( Figures 1 and S1 ) ., In summary , the null model allows mutation rates and levels of constraint to vary across CNCs , and it allows for the property that broad-scale mutation rates may vary across lineages ., In addition to the basic null model , we consider a family of alternative models that allow additional rate parameters for particular CNCs ., In the simplest alternative , a single branch on the tree evolves at a rate that is different from the background rate shared by the remaining lineages ( as for the third tree in Figure 1 ) ., In the extreme alternative , each of the seven branches evolves with its own rate ri ( h ) , giving a total of seven rate parameters for the CNC in question ., ( For simplicity of notation , we will henceforth drop the notation h on the rate parameters . ), In the extreme case , to test the hypotheses H0: r1 = r2 =, ···= r7 ( = r0 ) versus HA: r1 ≠ r2 ≠, ···≠ r7 at a particular CNC , we compute the SRT as, where L is the likelihood of the sequence data for the five mammalian species , maximized with respect to the rate parameters , and with the fixed estimate of branch lengths parameters (, ) and the sequence evolution model ., Large values of the SRT indicate a substantially better fit of the alternative than the null model ., Another example of alternative model is the case in which branches 2 and 3 have distinct rates r2 and r3 , while the other branches have a single “background” rate r0 , −2 , −3 ., In this case , to test the hypotheses H0: r1 = r2 =, ···= r7 ( = r0 ) versus HA: r2 ≠ r3 ≠ r1 = r4 =, ···= r7 ( = r0 , −2 , −3 ) , we can compute the likelihood ratio statistic as, In this paper , we perform two kinds of analyses ., One analysis performs model selection using the SRT , while the other tests for individual branches with rate changes ., When testing for a rate change on the ith branch only , it is convenient to transform the likelihood ratio statistic as follows ., In this case , we will use special notation , denoted by SRTi:, where sign ( x ) = 1 if x > 0 and otherwise sign ( x ) = −1 ., Rewriting the SRT in this way provides the convenient property that SRTi > 0 implies that ri is larger than the background rate r0 , −i , and hence branch i shows a rate speed-up relative to the rest of the tree; conversely , SRTi < 0 implies a slow-down on branch i ., As a convention , when we subscript SRT by a character or number , it will represent the signed likelihood ratio statistic testing for rate changes on the indicated branch ., Otherwise , the notation SRT without subscripts will be used to indicate use of an unsigned test statistic , in the form of Equations 2 and 3 ., Our SRT is a likelihood ratio test and , as such , standard theory suggests that under the null hypothesis the test statistic should asymptotically follow a chi-square distribution with degrees of freedom equal to the difference in the number of estimated parameters between the constrained ( null ) and less-constrained ( alternative ) models ., Similarly , the signed root of this statistic for a one-dimensional parameter of interest is asymptotically standard normal ., Therefore , when the null hypothesis is true and the number of sites in a CNC is large enough , the unsigned SRT might be expected to follow the chi-square distribution with the degrees of freedom equal to the difference in the number of rate parameters between the two models ., For example there are six degrees of freedom in the global test ( Equation, 2 ) and two degrees of freedom in the example in Equation 3 ., Similarly , under the null , the signed test SRTi is constructed to have a standard normal distribution as the CNC size goes to infinity ., Our simulation studies show that the asymptotic theory is reasonably accurate for both versions of the test statistic , except in the cases in which the lineages tested for selection are relatively short and are expected to accumulate few substitutions ( namely , the human and chimpanzee lineages; Figure S3 ) ., Hence , to reduce computational burden , we calculate p-values using the asymptotic chi-square or normal approximations , except for tests on the human and chimpanzee branches for which , except where stated , we compute p-values based on the empirical null distribution in simulated data ( see Methods ) ., An additional consideration is that we do not want the estimated null branch lengths ( vb ) to be heavily influenced by outlier CNCs with evidence for selection ., To mitigate the impact of such CNCs , we first identify CNCs with clear overall departures from the null model ( SRT > 25 in the global six degrees of freedom test , corresponding to p < 0 . 00034 ) , and then reestimate the branch lengths after dropping those nonneutral CNCs , which represent 2 . 8% and 3 . 8% of the total mammalian and amniotic CNCs , respectively ., In summary , then , our analysis performs the following steps: ( 1 ) Estimate maximum likelihood branch lengths and rates under the null; ( 2 ) identify outlier CNCs that have SRT > 25 comparing the seven- and one-parameter models; ( 3 ) drop outlier CNCs and recalculate the null branch lengths and rates; and ( 4 ) compute the shared rates test statistics for each CNC according to a range of alternative models ., For reasons discussed below , in practice these analyses were performed in a sliding window of 50 consecutive CNCs , as defined by position in the human physical map ., All analyses considered the mammalian and amniotic CNCs separately ., It is well established that the extent of divergence among mammalian species varies substantially across large genomic regions 33–38 ., For example , Gaffney and Keightley 38 showed that divergence between the mouse and rat genomes varied between and within chromosomes ., While the causes and the scales of this type of variation are not completely understood , it has been shown that divergence correlates with various genomic features , including GC and CpG content , simple-repeat structures , and recombination rate , suggesting that these genomic features drive variation in mutation rates 35 , 37 ., Variation in mutation rates or levels of CNC conservation across genomic regions should not be problematic for our method , provided that the substitution rate in any given region maintains a constant ratio to the average across the mammalian phylogeny ., If a CNC is in a region with a higher , or lower , mutation rate than average , this effect should simply be absorbed into the rate parameter that we estimate for each CNC as part of our null model ., However , if mutation rate variation is not stable across the phylogeny , this might produce false signals for our method ., Therefore , we looked at whether the average tree shapes are significantly variable across chromosomes ( according to the human physical map ) as well as within chromosomes ., We found that in fact there is nontrivial variation in tree shape , both at the chromosome level , and across genomic regions within chromosomes ., For example , within Chromosome 2 there is a highly significant autocorrelation in the fraction of the tree occupied by the mouse lineage ( Figure 2 ) ., This result implies that local variation in large-scale mutation rates is not conserved across evolutionary time; for example , genomic regions that evolve faster than average on some lineages may evolve slower than average elsewhere on the tree ., If average tree shapes were constant across the genome , we could use CNCs from across the genome to estimate the tree shape for our null model ., However , the observation that tree shape is not constant suggests that instead our model should allow for variation in tree shape across the genome ., After some experimentation , we settled on using a sliding window of 50 consecutive CNCs to estimate the tree shape ., That is , we test each CNC for significant departures from the tree shape in a 50-CNC window that , in the human physical map , is centered near the CNC in question ( see Methods ) ., On average , this window size corresponds to 525 kb and 1 . 3 Mb ( median ) for mammalian CNCs and amniotic CNCs , respectively ., Overall , we find that using the sliding window method produces only a modest impact on the rate of significant CNCs , but it should improve our inferences by taking into account the local variation in tree shapes ( Figures 2 and S4 ) ., An obvious concern about using a sliding window based on the locations of CNCs in humans is that due to chromosomal rearrangements , CNCs that are close together in humans may not be close together in other mammals ., Consequently , a sliding window based on the human map might not provide a suitable correction ., Fortunately , our window size is relatively small compared to the typical size of syntenic blocks 8 , 39 and in Figure 3 , we show that the results of tests on the human lineage are highly concordant whether we use windows based on the human or mouse physical maps and , indeed , are only modestly different from the results using all CNCs together ., Consequently , all subsequent results use 50-CNC windows based on the human map ., Another plausible concern about our model stems from the prediction that selection against weakly deleterious mutations is more efficient in species with large populations than in small populations ., This means that weakly constrained sites in CNCs are likely to evolve more quickly in primates than in rodents ( which have larger effective population sizes ) ., This effect has been observed in a comparison between the evolutionary rates of CNCs and putatively neutral flanking sequences 29 ., Hence—in contrast to our null model—one might expect the overall tree shape for a CNC to depend on its level of selective constraint ., To investigate this issue , we classified CNCs into four different levels of conservation , according to their substitution rates on the dog lineage ., We then separately compared the average human-to-chimpanzee divergence against the average mouse-to-rat divergence , within each of the four conservation levels ( Table S15 ) ., We find that that as the level of constraint increases , the divergence in rodents indeed decreases faster than divergence in hominids , consistent with the results of Keightley et al . 29 ., However , we find that the variation across CNCs is relatively small ( less than 11% change across different classes of CNCs ) and much less than when CNCs are compared to neutral sequences ( Table S3 ) ., As shown below , we do not have the power to detect such small variations in tree shape at individual CNCs , so we conclude that it is not necessary to control for overall conservation level more carefully for the current study ., For each CNC , we calculated SRTi for each of the seven branches of the mammalian tree to identify CNCs that have experienced a speed-up or slow-down on a particular branch ., Figure 4A shows the histogram of p-values on the mouse lineage ( SRTm ) for the mammalian CNCs ., The p-values are defined as P ( SRTi > srti ) where srti is the observed value ., Hence , p-values near 0 indicate increased rates , and near 1 indicate decreased rates ., The histogram is flat for intermediate p-values with peaks at both ends , suggesting that most CNCs fit the null distribution of SRTm , but with a substantial number of outliers ., At the significance level of 0 . 001 , 1027 ( 1 . 2% ) and 503 ( 0 . 6% ) mammalian CNCs show speed-ups and slow-downs , respectively ., Among amniotic CNCs , 228 ( 1 . 4% ) and 106 ( 0 . 6% ) show speed-ups and slow-downs , respectively on the mouse lineage ., Figure 4B plots the expected and observed branch lengths on the mouse lineage for the CNCs that are significant at p < 0 . 001 in each tail ., ( Similar plots for other lineages are shown in Figure S5 . ), The red points above the diagonal indicate CNCs with rate speed-ups ., For the central 95% of the significantly fast-evolving CNCs , the observed branch lengths are between 0 . 04 to 0 . 13 substitutions per site , and are 2–4-fold higher than the expected branch lengths ., The blue points below the diagonal are CNCs with reduced branch lengths ., Nearly half of these CNCs accumulated no substitutions on the mouse lineage ., The other long lineages show similar p-value histograms though with some variability in the proportion of significant CNCs ., The dog lineage is the most enriched for signals , with 2 . 3% and 1 . 9% of mammalian CNCs showing speed-ups and slow-downs , respectively , at p < 0 . 001 ( in each tail ) ., Even after a stringent Bonferroni correction , 186 and 46 CNCs , respectively , are still significant at p = 0 . 001 in the dog lineage ., The overall results for amniotic CNCs are similar , but the fraction of significant results is slightly higher on each branch ( Table S8 ) ., For most lineages , our significance threshold ( one-sided p-value < 0 . 001 on each end ) corresponds to a genome-wide false discovery rate ( FDR ) between 0 . 05 and 0 . 1 ( Table S9 ) ., Since the distribution of SRTi on the human and chimpanzee lineages does not follow the standard asymptotic distribution , we simulated data under the null over a range of substitution rates that cover the observed range over all 50-CNC windows ( see Methods ) ., We account for heterogeneity in the distribution of SRTi across bins of CNCs with different numbers of expected substitutions on the tested lineage by computing p-values based on the empirical null distribution of SRTi constructed in each bin ( unpublished data ) ., At a significance level of 0 . 001 , 256 mammalian CNCs and 59 amniotic CNCs , respectively , show rate speed-ups on the human lineage ( Table S8 ) ., Note that there is little power to detect rate reductions on these very short lineages ., To better understand these SRTi results , we performed power simulations under a range of models ., The simulation results , summarized in Figure S6 , show considerably greater power to detect speed-ups than slow-downs on all lineages , consistent with the results of Siepel et al . 40 ., Thus , the fact that we detect more speed-ups than slow-downs does not necessarily imply that speed-ups are actually more common , and it is likely that many slow-down events are simply not detected by our analysis ., Our human results allow a comparison to the human accelerated regions ( HARs ) identified by Pollard et al . 5 using a similar type of approach , based on regions that were highly conserved ( at least 96% identity ) across chimpanzee , mouse , and rat ., Among the top 49 HARs , which include two coding regions , 34 overlap with CNCs in our dataset; however , generally , the HARs are considerably shorter and more conserved and lie within our CNCs ., Perhaps not surprisingly , since the HARs are the top genome-wide hits in their data , the signals in our overlapping CNCs tend to be weaker ., Among the 34 CNCs , just five CNCs are significant in our analysis at a genome-wide FDR less than 0 . 05 ., Nonetheless , our CNCs that overlap HARs do show a strong enrichment of modest signals ., Our human lineage p-values are <0 . 01 for 26 of the 34 CNCs overlapping HARs , and are <0 . 1 for 33 of the 34 ( Table S10 ) ., Within our dataset , one of the most significant CNCs on the human lineage is a 144-bp amniotic CNC located on human Chromosome 21 starting at 33481809 ( q22 . 11 , NCBI Build 35 ) ., It was not detected by Pollard et al . 5 because it fails their filtering threshold for similarity between chimpanzee , mouse , and rat ., As illustrated in Figure 5 , the posterior expected number of substitutions ( see Methods for details ) on the human lineage is 5 . 2 , which is 26-fold higher than the value of 0 . 2 expected under the null model ., The corresponding SRTh is 4 . 84 ., The p-value for this CNC is so small that it is difficult to evaluate by simulation; however , the standard normal approximation suggests that p ≈ 6 × 10−7 ( our simulations indicate that this is conservative ) ., In addition to the five nucleotide substitutions , there is also a 2-bp insertion on the human lineage that was not included in the statistical inference ., Since the UCSC genome browser database was recently updated , we were able to inspect an alignment of 17 vertebrate species for this region ., Manual inspection confirmed that all six of these substitutions occurred on the human lineage ., The function of this CNC is unclear but the two nearest genes are C21orf54 , 17 kb upstream , and IFNAR2 , 42 kb downstream of the CNC ., Not much is known about C21orf54 , but IFNAR2 codes for a type I membrane protein that forms one of the two chains of a receptor for interferons alpha and beta 41 ., This CNC is strongly conserved among the other mammalian species and chicken but does not appear to be present in the fugu genome ., In addition to the rapid evolution on the human lineage , there is weak evidence for slower evolution of this CNC on the mouse and dog lineages ( one-sided p-values = 0 . 011 and 0 . 023 , respectively; see Figure 5B ) ., Thus far , we have focused on the simplest class of alternative models , in which a CNC changes substitution rate on a single branch only and has a constant background rate elsewhere on the tree ., We now extend this approach in order to classify each CNC according to a family of more complicated models of evolutionary patterns ., Our data are connected by a tree containing seven branches ., The simplest model ( our “null” ) has a single rate parameter , and the most complicated alternative model has seven different rate parameters ., In between , there are 876 ways of partitioning the seven branches into two or more different substitution rate groups ., However , considering all of these partitions does not seem biologically meaningful or necessary , and here we focus on a reduced set of 126 alternative candidate models ., The alternative models we consider can be divided into two distinct classes of models ., In one class of models , each tree is assumed to have a “background” rate parameter ., Then , each CNC may have between one and six “selected” lineages , and each selected lineage evolves at its own rate ., In the other class of models , each tree may be split into subtrees that share a single rate , while the rest of the tree has a single background rate ( for full details , see Table S11 ) ., We use a modified Akaike Information Criterion ( AIC ) procedure to classify each CNC into its best model ., In brief , the method attempts to account for multiplicities of alternative models as well as the number of estimable parameters in each model ( see Methods ) ., We have performed simulations to test the performance of this method , and we find that it provides suitable control over the rate of “false positives” ( i . e . , accepting models with more parameters than used to simulate the data ) ., That said , our simulations show that it is often difficult to correctly classify complex models with multiple rate changes ( see Methods; Figure S7 ) ., The results of our data analysis are summarized in Figure 6 and Table S12 ., We estimate that ∼68% ( 54 , 643/81 , 957 ) of the mammalian CNCs evolve at a single rate ., The remaining nonneutral CNCs show rate changes on at least one lineage ., The number of CNCs assigned to each model category decreases with increasing model complexity ., Among the 32% of CNCs with more than one rate , ∼75% ( 20 , 420/27 , 314 ) exhibit rate changes on a single lineage but not on the remaining lineages and ∼9% ( 2 , 419/27 , 314 ) exhibit rate changes on the primate or the rodent lineage that are inherited across all branches below ., For the two-parameter models , the rate change events are easily classified as speed-ups or slow-downs ., Counts for both types of event are shown in Figure 6B ., For most lineages , there are slightly more speed-up events than slow-downs ( ∼55% versus ∼45% ) ., However , there are 638 and 530 CNCs that show rate speed-ups on the human and chimpanzee lineages , respectively , far more than the four and eight CNCs , respectively , showing slow-downs ., Presumably , these results are due in large part to the greater power to detect speed-ups , as well as differences in power across lineages ( Figure S6 ) ., It is notable that the dog lineage shows a very large number of rate changes , which may not be fully explained by the long length of this lineage ( second longest among the seven ) ., Since there is no strong tendency towards an excess of speed-ups over slow-downs on this lineage , it is unlikely that this can be explained by occasional CNCs with low-quality dog sequence ., Perhaps a hint is that we have observed greater variation in the dog-lineage substitution rates at neutral sites than on other lineages ., Perhaps there is greater fine-scale variation on the dog lineage that is not well captured by our 50-CNC window method ( see Methods; Figure S8 ) ., As discussed above , we have identified many CNCs with significantly accelerated rates on one or more branches ., However , it is unclear a priori whether these speed-ups reflect positive adaptation or relaxation of functional constraint ., In order to address this issue , we estimated substitution rates in unconserved sequences near each CNC to estimate local neutral rates ( see Methods ) ., We then determined how many of the CNCs showing rate speed-ups have an accelerated rate that actually exceeds the corresponding lineage-specific neutral rate ., If the rate in a CNC actually exceeds the local neutral rate , this is strong evidence for adaptive evolution ., However , a negative result here is difficult to interpret , since adaptive evolution in an otherwise slow-evolving sequence may not necessarily bring the total rate above the neutral background rate ., Our results are summarized in Table 1 ., We observe that most CNCs showing accelerations on the human and chimpanzee branches indeed have rate estimates exceeding the neutral rates; of these , more than half are actually significantly faster than the neutral rate at p < 0 . 05 ., Meanwhile , the other branches of the mammalian tree all show smaller fractions of CNCs with rates that exceed the neutral rate , and very few of these are significantly faster than the neutral rate ., One plausible explanation might be that if there is sufficiently rapid evolution on a long branch , this might cause an otherwise conserved element not to be classified as a “most conserved” region by the HMM 10 ., However , some simple calculations suggest that this is likely to be a modest effect in practice ., Moreover , we see the same effect for both the mammalian and amniotic CNCs ( Table 1 ) , even though the HMM data for the latter include the relatively long branch to chicken , and should therefore be much less susceptible to this effect ., Instead , to explain these observations , we hypothesize that the rate speed-ups that we detect may often reflect rapid bursts of adaptation in which a CNC accumulates a series of sequence changes , thus modifying its function ., A single burst of adaptation may produce enough sequence changes to exceed the neutral rate on a short branch , but not on a longer branch ., In this model , we would have the most power to detect adaptive events on short branches ., Our data argue strongly against a model in which a CNC adapts continuously over extended periods of evolutionary time , as such a model should also produce signals on the long branches ., We have also performed analyses of the locations of CNCs showing branch-specific speed-ups , with respect to nearby genes ., A recent report by Drake et al . 27 found that the frequency spectrum in CNCs is most skewed towards rare variants ( indicating weak purifying selection ) in introns and near genes , and is less skewed in CNCs that are far from genes ., To test whether CNCs showing speed-ups on particular branches occur at higher rates near to or far from genes , we divided all our CNCs into four classes: intronic , within 10 kb of a gene , between 10 kb and 100 kb , and greater than 100 kb from any gene ., We found that on the mouse and rat lineages , CNCs showing speed-ups ( p < 0 . 001 on the branch-specific test SRTi ) occur at higher rates in introns and within 10 kb of genes than among CNCs further from genes ., However , this trend was not replicated on the other lineages of the tree ( Table S13 ) ., We next looked at whether CNCs showing significant rate speed-ups are more likely to be in the proximity of particular kinds of genes 17 , using the PANTHER GO database 32 ., A significant difficulty in this sort of analysis is that even for those CNCs that act as cis-regulators , it is unknown which of the nearby genes is being regulated ., However , as a rather imperfect proxy for this we simply used , for each CNC , the nearest gene ( in either orientation ) ., For each branch of the mammalian tree , we divided the CNCs into those with increased rate on that branch ( by AIC ) and used CNCs evolving under the null model as “neutral” controls ., We looked at whether particular biological process categories were enriched among the nearest genes of the selected CNCs compared to the neutral CNCs ., For mammalian CNCs , there is significant enrichment of the process categories “amino acid activation” and “other coenzyme and prosthetic group metabolism” on the dog and the lineage leading to the common anc

Introduction, Results, Discussion, Methods, Supporting Information

Conserved noncoding elements ( CNCs ) are an abundant feature of vertebrate genomes ., Some CNCs have been shown to act as cis-regulatory modules , but the function of most CNCs remains unclear ., To study the evolution of CNCs , we have developed a statistical method called the “shared rates test” to identify CNCs that show significant variation in substitution rates across branches of a phylogenetic tree ., We report an application of this method to alignments of 98 , 910 CNCs from the human , chimpanzee , dog , mouse , and rat genomes ., We find that ∼68% of CNCs evolve according to a null model where , for each CNC , a single parameter models the level of constraint acting throughout the phylogeny linking these five species ., The remaining ∼32% of CNCs show departures from the basic model including speed-ups and slow-downs on particular branches and occasionally multiple rate changes on different branches ., We find that a subset of the significant CNCs have evolved significantly faster than the local neutral rate on a particular branch , providing strong evidence for adaptive evolution in these CNCs ., The distribution of these signals on the phylogeny suggests that adaptive evolution of CNCs occurs in occasional short bursts of evolution ., Our analyses suggest a large set of promising targets for future functional studies of adaptation .

Conservation of DNA sequences across evolutionary history is a highly informative signal for identifying regions with important biological functions ., In particular , conserved noncoding regions have been shown to be good candidates for containing regulatory elements that have roles in gene regulation ., Recent studies have found that there are many thousands of conserved noncoding elements ( CNCs ) in vertebrate genomes and have suggested possible functions for some of these elements , but the function of most CNCs remains unknown ., To study the evolution of CNCs , we developed a statistical method to identify CNCs that show changes in evolutionary rates on particular branches of the mammalian phylogenetic tree ., Those rate changes may indicate changes in the function of a CNC ., We applied our method to CNCs of five mammalian genomes , and found that , indeed , many CNCs have experienced rate changes during their evolution ., We also found a subset of CNCs showing accelerations in evolutionary rate that actually exceed the neutral rates , suggesting that adaptive evolution has shaped the evolution of those elements .

evolutionary biology, genetics and genomics, mammals, computational biology

journal.pgen.1002236

A Comprehensive Map of Mobile Element Insertion Polymorphisms in Humans

Retrotransposons are endogenous genomic sequences that copy and paste into locations throughout host genomes 1–3 ., Most mobile elements annotated in the human reference genome are remnants of ancient retrotransposition events and are no longer capable of active retrotransposition ., However , a fraction of mobile elements remain active and contribute to variation between individuals in the human population ., These active elements belong almost exclusively to the Alu , L1 , and SVA families of non-LTR retrotransposons 4 ., The Alu family is the most common mobile element in primate genomes , with more than 1 . 1 million copies in Homo sapiens 5–7 ., The sequence of a full-length Alu element is 300 bp long ., Alu elements are classified into a range of sub-families which have different propensities for retrotransposition , and are identified according to sequence alterations ., Several AluY sub-families are currently active and are responsible for the bulk of mobile element insertion variation in Homo sapiens ., The human reference genome contains over 140 , 000 annotated AluY elements ., After Alus , L1 insertions are the next most prevalent family of mobile elements ., There are over 500 , 000 L1 elements annotated in Homo sapiens ., A full-length L1 is a sequence of roughly 6 kb in length and the most active L1 sub-family in the human lineage is L1HS 8 , 9 ., There are a little more than 1 , 500 L1HS annotated elements in the human reference ., A third family of mobile element are SVA retrotransposons 10 ., SVAs are hybrid elements of SINE , VNTR and Alu components that range in size up to several Kb , with more than 3 , 600 annotated SVA elements in the human reference genome ., SVA elements are thought to be the youngest family of retrotransposons in primates 11 ., Other less common classes of mobile elements , such as DNA transposons , and endogenous retroviruses are not the focus in this study ., Mobile element insertions ( MEI ) are known to generate significant structural variation within Homo sapiens 12 , 13 and have diverse functional impacts 14–16 ., In vitro experiments identified key features of Alu 17 and L1 18 elements responsible for retrotransposon activity ., The identification of MEI variant loci in humans initially began with disease-causing insertion events ( e . g . hemophilia 19 , breast cancer 20 ) ., Experimental approaches were based upon library screening and small-scale PCR based display assays 21 ., These approaches have been augmented by comparisons of the NCBI and the HuRef genomes 22 , 23 , large scale fosmid-end sequences 24 , and targeted sequencing of element-specific PCR products 25–28 ., The dbRIP database of MEI polymorphisms 29 currently contains 2 , 691 polymorphic loci , enabling early estimates for the total number of segregating events 25 and per-generation mutation rates 23 ., MEI polymorphisms can be detected either as insertions or as deletions in samples relative to the reference genome ., Mechanistically , however , both types of observations are due to retrotransposon insertion; precise excisions of mobile elements are essentially non-existent 1 ., Therefore MEI detected as deletions are , in fact , retrotransposon insertions in the reference DNA and can be verified as such by comparison with ancestral genomes ., Detection and genotyping properties of MEI detected as insertions ( “non-reference MEI” ) and as deletions ( “reference MEI” ) are substantially different ., We present their respective properties separately before combining the two detection modes into a unified MEI analysis ., The deletion detection methods and properties of the full set of 1000GP deletions have been extensively described in the 1000GP CNV companion paper 30 ., This allows us to focus on specific properties of the reference MEI subset of those deletions ., Effective computational algorithms using second-generation sequencing data exist for identifying deletions 27 , 31 , 32 , and have been used to find MEI in particular 33 ., Detecting non-reference MEI directly as insertions from whole genome shotgun sequence data poses a more challenging problem , owing to the inherent difficulties associated with accurate mapping of sequenced reads derived from highly repetitive regions of the genome ., Only recently have methods been developed for the purpose of non-reference MEI detection from second-generation whole genome shotgun data including published studies of L1 element insertions 34 and of Alu insertions 35 ., These studies adopted similar computational approaches to one of our insertion detection methods ( the read pair method , see Materials and Methods ) and have different detection properties ( Text S2 Comparisons , Figures S8 , S9 , S10 ) ., Relative to previous studies , we present a broad analysis of MEI variation in the human population; with more variant loci detected , from the three major mobile element families , using multiple detection methods , each with comprehensive experimental validation ( Table 1 ) ., The present study represents the combined efforts of the MEI sub-group of the 1000 Genomes Project and has been prepared as a companion to previous 1000GP pilot publications 30 , 36 ., The MEI analyzed in this study were included the 1000GP variant call release of July 2010 ( ftp://ftp-trace . ncbi . nih . gov/1000genomes/ftp/pilot_data/release/2010_07 ) , also provided as Table S1 ) ., The specific purpose here is to provide a more detailed description of the methods , validation experiments , and properties of the 1000GP catalog of MEI events , and to extend the analysis by adding genotype information , population allele frequencies , and population specific mutation rates ., We analyzed two whole-genome datasets produced by the 1000GP , the low coverage pilot dataset consisting of 179 individuals sequenced to ∼1–3X coverage and the trio pilot dataset consisting of two family trios sequenced to high , ∼15–40X coverage ( Table S2 , Figure S4 ) ., These datasets included samples from three continental population groups , 60 samples of European origin ( CEU ) , 59 African ( YRI ) , and 60 Asian samples from Japan and China ( CHBJPT ) ., The two pilot datasets were produced and analyzed for complementary purposes ., The trio dataset was used for assessing detection methods in high coverage samples and for the purpose of finding candidate de novo insertions in the trio children ., The high coverage dataset was used to assess population properties of MEI ., Both datasets contributed to the overall catalog of events ., We developed two complementary methods for the detection of non-reference MEI , a read-pair constraint ( RP ) method applied to Illumina paired-end short read data , and a split-read ( SR ) method applied to the longer read data from Roche/454 pyrosequencing ( Materials and Methods: non-reference MEI detection ) ., Figure 1a and 1b shows the respective detection signatures and examples of event displays ., Candidate MEI events were formed as clusters of supporting fragments ., A limitation specific to RP detection arises from annotated elements within a characteristic read pair fragment length of candidate MEI ( Figure 1a ) ., Read pairs spanning from a uniquely mapped anchor into an annotated mobile element with a fragment length consistent with the given library fragment length distribution ( Figure S5 ) are characteristic of the reference allele and are not evidence for non-reference MEI ., These “background” read pairs occasionally have fragment lengths on the extreme tails of the library distribution and can potentially be misclassified as evidence for non-reference MEI ., For this reason RP detection criteria required at least two supporting fragments spanning into the insertions from both sides of the insertion ., We also masked insertion positions within a fragment length around each annotated element of the corresponding family from RP detection in order to achieve a low false detection rate ., The SR method was not dependent on the fragment length distribution in the 454 data so these additional detection criteria were not required ., We applied the two methods to both 1000GP pilot datasets ( Table 1 ) separately , yielding a total of 5 , 370 distinct genomic MEI loci , 33% of which were found by both SR and RP methods ( Figure 1c ) ., The overall level of detection overlap between SR and RP methods is limited by detection sensitivity and specificity ( see below ) and the number of samples sequenced by both 454 and by Illumina read pairs ., In addition to the 5 , 370 non-reference MEI , we identified 2 , 010 reference MEI detected as deletions of mobile elements in samples ., The reference MEI events were selected from the full release set of 1000GP pilot deletions ( n\u200a=\u200a22025 ) 30 , 36 based on matching deletion coordinates to RepeatMasker 3 . 27 Alu , L1 , and SVA annotations 6 , and the requirement that the mobile element is absent in the chimpanzee genome 37 ( 6x pan Trogodytes-2 . 1 assembly ) at the corresponding positions in hg18 ( Materials and Methods: Reference MEI selection ) ., Figure 1e shows an example event display of an AluYb8 reference MEI , detected as a deletion in the trio pilot data ., All but one of the reference MEI were found by one or more of the RP or SR deletion detection algorithms that were part of the released 1000GP deletion call set 30 , 38–42 with a small overlapping contribution from algorithms based on assembly or read depth methods 43 , 44 ( Figure 1d , Table S3 ) ., The complete set of 7 , 310 MEI calls is simply the combined set of reference and non-reference MEI over both pilot datasets ( summarized in Table 1 , complete list in Table S1 ) ., Insertions occurring at the same locus from different call sets were merged using a 100 bp window for matching positions , choosing the SR insertion coordinate when available to represent the merged event ., Similarly for reference MEI , deletion merging was accomplished among the 23 separate 1000GP call sets using a precision-aware algorithm described in detail in the 1000GP SV companion paper 30 ., The full catalog of MEI loci appear to be distributed randomly across the genome ( Figure 2b ) with a characteristic spacing of 0 . 4 Mb between MEI loci , except for an apparent MEI hotspot in the HLA region of chromosome 6 where 19 MEI loci are clustered in a 1 Mb region ( 8 times the genomic average density for MEI , Figure S11 ) ., Accurate read mapping in the HLA region is complicated by a high density of variation 36 , however , we see no evidence of falsely detected MEI here ., The balance between reference and non-reference MEI , proportions of RP and SR detected loci , the fraction of previously identified MEI loci , and the validation rate are all consistent with genomic averages; only the density of MEI is significantly increased ., The genomic proportions of the three mobile element families are 85±2% Alu , 12±2% L1 , and 2 . 5±1% SVA ( Figure 2b ) for both reference and non-reference MEI ., Most non-reference MEI loci were detected from the low coverage pilot data ( Figure 2c ) while the reference MEI were more evenly distributed between the low coverage and trio pilot data ( Figure 2d ) ., As described in the 1000GP main pilot paper 36 , more than 80% of the non-reference MEI were newly identified loci not detected by previous studies 23–28 , 34 , 35 , 45 ., However , in the mean time , several published studies have produced new lists of non-reference MEI loci including L1 insertions 34 and Alu insertions 28 , 35 ., Half of the non-reference MEI loci from this study have not yet been reported elsewhere ( Figure 2e , Figure S8 ) ., Table 1 of the 1000GP paper lists 5 , 371 MEI , two of these events were subsequently merged into one to form the present count of 5 , 370 MEI detected as insertions ., For reference MEI , we find that 76% of our events matched deletion coordinates listed in the dbVAR ( 28 January 2011 ) structural variation database or a deletion identified in the HuRef genome 22 , 46 , leaving 24% of the reference MEI unreported prior to 1000GP publications ., The 1000GP catalog of MEI variant sites includes all 7 , 310 detected loci , including those matching MEI from other publications ., Further comparisons among the recent MEI studies are provided in Text S2 ., We benchmarked each of the four non-reference MEI call sets ( separate SR and RP call sets for the low coverage and trio pilot datasets ) to assess detection sensitivity and specificity ., As MEI are currently not suitable for microarray validation due to their highly repetitive sequence , all validations were done by locus-specific PCR ., 200 loci were randomly selected from each of the four insertion call sets ., Using an automated pipeline 32 , primer design was possible for 746 loci ( Table S4 ) ., In addition to the randomly selected loci , other candidate loci were selected for validation experiments in order to confirm SVA insertions ( n\u200a=\u200a7 ) , to test potential de novo insertions from the pilot 2 trio ( n\u200a=\u200a1 ) , and gene-interrupting events ( n\u200a=\u200a86 attempted ) , as well as for algorithm training and testing purposes ( n\u200a=\u200a386 ) ., These additional PCR results ( Table S4a ) were not used to assess false detection rates , except for the special case of SVA insertions , which were under-represented in the random loci selection since SVA insertions are relatively rare ., All candidate loci with successful primer design were tested on two different population genetic panels ( Materials and Methods: Validation ) one with DNA of 25 individuals from the low coverage pilot , and one with DNA from all samples of the trio pilot dataset ., In addition to other human samples from populations not represented by the pilot datasets , DNA of a chimpanzee was also included on the panel to confirm that the identified insertion is indeed human-specific ., An example of typical results for a low coverage locus is shown in Figure 3a ., Through additional primer design for loci with inconclusive results and PCRs using a primer residing within the 3′ end of a retrotransposon , in particular within SVA elements , more than 98% of the tested candidate loci were successfully genotyped ., The validation experiments revealed overall insertion false discovery rates for each dataset of less than 5% ( Table 1 ) ., Among the different retrotransposon families ( L1 , SVA , and Alu elements ) , false discovery rates varied noticeably ( Figure 3b ) , with Alu insertions showing the lowest false-positive rate ( 2 . 0 1 . 1–3 . 4 % , followed by L1s ( 17 10–27 % ) , and SVAs ( 27 8–55 % ) with 95% confidence intervals ., This is not entirely unexpected as polymorphic Alu insertions tend to be low divergence full-length AluY elements , unlike L1 or SVA insertions which tend to be truncated and may be accompanied by adjacent transduced genomic DNA sequences ., Although the SR and RP detection methods are very different , the overall detection specificities were remarkably consistent ., Following the validation of non-reference MEI , we assessed detection sensitivity ., The primary challenge here was to find suitable gold standard non-reference MEI that should be present in our samples from which to assess sensitivity ., We estimated sensitivity in three different ways , as a consistency check ., First , we estimated sensitivity by using the high quality non-reference MEI from HuRef 23 as a gold standard and found that 74% of the 650 Alu , L1 , or SVA insertions in HuRef matched MEI insertion loci in our catalog ( Table S5 ) ., This represents a lower limit for insertion detection sensitivity since not all MEI in the HuRef genome are necessarily present in the 1000GP pilot samples ., Next we looked at the overlapping insertion detection between the RP and SR methods in the trio children samples ( Figure 3c , Figure S6 ) , which were the samples sequenced to the highest depth for both Illumina and 454 data ., Based on the detected loci overlap ( see Materials and Methods: Detection sensitivity ) , we estimate 67%±3% and 70%±7% sensitivities respectively for RP and SR insertion detection in the trio children ( Table S6 ) , with a combined SR+RP detection sensitivity exceeding 90% in the CEU trio child ( see Materials and Methods , Eq . 4 ) with high coverage data from both 454 and Illumina reads ., A third approach to estimate for the non-reference MEI detection sensitivity is based on the validation PCR genotypes in the low coverage dataset ., Since the PCR loci were selected as random subsets for each RP and SR call set independently , the validated sites selected from SR events can be used as a gold standard to assess RP detection sensitivity , and vice-versa ., Detection sensitivity as a function of allele frequency ( Figure 3d ) was estimated for each method from PCR genotypes at those loci randomly selected for validation of the complementary method ., PCR genotypes provided the allele frequency estimate on the abscissa ., Statistical errors at high allele frequency are large because the limited number of tested MEI loci at higher allele frequencies ., Detection sensitivity of the RP method saturates close to 70% at high coverage and the SR method sensitivity exceeds 70% at high coverage ( Figure S6 ) ., The corresponding trend is apparent in Figure 3d ., The combined detection sensitivity approaches 90% for common alleles ( Materials and Methods , Eq . 4 ) ., However , since relatively few of the low coverage samples were sequenced with 454 , a realistic estimate for the detection sensitivity to common MEI insertions is between 70% and 80% ., This is consistent with 75% derived from the HuRef gold standard comparison and the sensitivity estimate from the trio pilot overlaps ., Equivalent estimates for Alu , L1 , and SVA specific sensitivities for common MEI alleles are 75%±10% , 50%±10% , and 50%±20% respectively ( Table S9 ) ., Regarding reference MEI detected as deletions , the overall validation rate from PCR and local assembly for the MEI component of deletions was 96% ., This does not imply that the remaining 4% were false , only that the released set of deletions contained reference MEI detected by two high specificity algorithms with characteristic false detection rates less than 10% ., These algorithms did not require additional validation evidence in the 1000GP release ., A rough estimate for the false detection rate for the MEI component of deletions is therefore 0 . 4% ., The number of algorithms supporting a given call is another indicator of call quality ., The average number of separate deletion calls ( out of a maximum of 23 call sets ) supporting events in the MEI subset was 7 . 8 while the average over all other deletions was 2 . 3 ( Figure S2 ) ., The high validation rate and high consensus among detection algorithms indicate that this subset of deletions is relatively free of detection artifact ., The practical limitation on the specificity of these events as reference MEI is the subsequent MEI selection criteria ., Only a small fraction the 2 , 010 selected events were ambiguous in terms of matching coordinates to an annotated mobile element with corresponding gap in the chimpanzee genome assembly ( e . g . Figure S3 , bottom panel ) ., The 1000GP CNV paper identified 2029 reference MEI variants using the BreakSeq algorithm ., Overlap between the respective lists is 89% ., We estimate 10% as an upper limit on the false discovery rate for reference MEI ., Detection sensitivity for reference MEI was estimated from the fractions of gold standard reference MEI identified by Xing et al . from HuRef 22 , 23 , 46 , and reference MEI identified by Mills et al . 4 , 47 from 1000GP samples NA12878 and NA12156 matched to any of our 2 , 010 reference MEI ( Table S5 ) ., In each case the fraction of those MEI deletions found in this study exceeded 90% ., This level of detection sensitivity is considerably higher than the bulk deletion detection sensitivity reported in the SV companion paper 30 , indicating that the RP and SR deletion detection methods developed for the 1000GP were particularly well suited for reference MEI detection ., We characterized each detected MEI event ( Table S1 ) by the insertion position , which algorithm ( s ) detected the event , number of fragments supporting the insertion and reference alleles , insertion length ( Figure S12 ) , element family , bracketing homology ( Figure 4a ) , and assembled sequence ., MEI have a characteristic “target site duplication” region of homology bracketing the insertion ., The target site duplication length distributions for the MEI detected by different methods , as well as for different element families , peaked at 15 bp with a standard deviation of 7 bp ( Figure 4a ) ., The full insertion sequence from reference MEI is readily extracted from the reference , but non-reference MEI require local de novo assembly to reconstruct the inserted sequence ., For this we used 454 data to reconstruct 1 , 105 Alu insertions ( Tables S1 and S7 ) from our event list based on the PHRAP assembly program 48 ., We then used BLAT 49 to map assembled contigs back to the build 36 . 3 human reference to identify the boundaries of the inserted sequence ., The inserted sequence was then mapped back to the RepeatMasker mobile element sequences using the RepeatMasker web server ( http://www . repeatmasker . org ) to identify the sub-family ( Figure 4b ) ., The accuracy of Alu sub-family classification was assessed by comparison to matched 359 Alu insertions from dbRIP 29 and nine fully sequenced Alu insertions from PCR validation experiments ., 272 of the assembled Alu sub-family classes were identical ( 74% ) ., The most active Alu sub-families are AluYa5 and AluYb8 ., AluY sub-families account for essentially all Alu variation ., The relative proportions among Alu sub-families are consistent among reference and non-reference MEI , as well as consistent with the Alu sub-families observed in HuRef 23 ., The Alu sub-family breakdown differs from that reported by Hormozdiari et . al . 35 who identified more than 10% of their set of insertions from AluJ or AluS sub-families ., The authors of that study point out that these ‘older’ Alu events could arise from mechanisms other than retrotransposon insertions ., Genotyping of non-reference MEI ( Materials and Methods: Genotyping ) was based on counts of fragments supporting the reference allele and fragments supporting the insertion allele at each locus for each sample ., Heterozygous MEI sites are identified by roughly equal amounts of reference and alternate allele supporting fragments spanning an insertion locus , while homozygous sites should have all fragments supporting one or the other allele ., For reference MEI , we used genotypes produced by the Genome STRiP package 39 , which was developed for 1000GP deletion genotyping 30 , 39 and incorporates Beagle 50 imputation based on linkage with local SNPs ., Both genotyping methods provide phred-scaled 51 genotype quality ( GQ ) metrics at each site that reflect confidence in the given call based on supporting evidence , GQ\u200a=\u200a0 to a total lack of genotype evidence and GQ\u200a=\u200a10 indicating that the genotype should be 90% accurate ., The GQ metric depends on the number of fragments found to support the MEI and non-MEI alleles for a given locus and sample ( Text S2: Genotyping methods ) ., As in most issues of sensitivity vs . specificity , there is a trade-off between high genotype efficiency and genotyping accuracy ., The drop-off in genotyping efficiency vs . GQ threshold is more severe for non-reference MEI ( Figure S13 ) ., For subsequent genotype-based analysis of non-reference MEI sites and samples we required GQ≥7 , which corresponds to roughly 40% genotyping efficiency in the low coverage pilot data ., For reference MEI we required GQ≥10 , which corresponds to an efficiency of 80% ., Genotyping efficiency improves with increased sample read coverage ( Figure S13 , bottom panel ) , particularly for non-reference MEI ., Genotyping accuracy for non-reference MEI is assessed by direct comparison to PCR validation genotypes in the same samples , and by testing for Mendelian errors in the trios and violations of Hardy-Weinberg Equilibrium in the low coverage data ( Text S2 Genotyping tests , Figures S13 and S14 ) ., Validation genotypes are listed in Table S4 ( also as the “VG” field of the released MEI insertion genotyped VCF files ) ., Genotype contingency tables for the low coverage data ( Table 3 ) show an 87% agreement between sequenced genotypes and PCR genotypes for sites with GQ≥7 ., Genotyping accuracy improves with increasing GQ threshold ( Figure S13 ) but never exceeds 90% in the low coverage data ., Non-reference MEI genotyping performance for high coverage trio data ( Table 3 , Table S8 ) was considerably better than for the low coverage data ., However , for population analyses we used only low coverage data in order to minimize the potential for coverage biases ., The accuracy of GenomeSTRiP genotypes ( for reference MEI events ) with GQ≥10 was estimated at 99% in the full 1000GP deletion call set 30 , 36 , 39 ., We estimated MEI allele frequencies from the count of high quality ( GQ≥7 non-reference and GQ≥10 for reference MEI ) genotyped insertion alleles for each MEI locus ., Allele frequencies were estimated from loci with at least 25 high quality genotypes for each continental population group ., The two MEI detection modes ( i . e . reference and non-reference insertions ) have very different allele frequency spectra ( Figure 5a–5c ) ., Since the non-reference MEI and reference MEI components have very different powers of detection and genotyping , the two components were corrected separately ( Materials and Methods: Allele frequency spectra ) before being combined into the full MEI spectrum ( Figure 5d–5f ) ., We estimated correction factors for each population group , each element type , and each detection mode ( Table S9 ) ., Non-reference MEI correction factors are larger than reference MEI factors because of the lower detection sensitivity and genotyping efficiency ., The allele count spectra were compared to the standard neutral model 52–54 , θ/i , where θ is an MEI diversity parameter and i is the allele count in a fixed number of samples ., The value of θ is fit from the MEI allele count spectrum for each population group and the fitted model is the gray dotted line appearing in Figure 5d–5f ., Only allele count bins in the range 0 . 15<frequency<0 . 95 were used in the fit ( bins marked with error bars in Figure 5d–5f ) to avoid regions of poor detection sensitivity ., The corresponding gray dashed lines superimposed on Figures 5a–5c also represent the neutral model expectation , modified to account for the respective ascertainment conditions , ( θ/2N ) for reference MEI , ( θ/i ) ( 2N−i ) / ( 2N ) for non-reference MEI , where N\u200a=\u200a25 is the number of samples in the spectra ., These ascertainment condition expressions are based on the assumption that the reference genome represents a random sample from the given population , which is admittedly simplistic but nevertheless explains much of the difference between the allele spectra of reference and non-reference MEI ., A coalescent simulation ( Text S2 Coalescent , Figure S17 ) for MEI variation also shows this behavior using standard population history parameters 55 ., Fitted values of the diversity parameter θ for each of three population groups and each element family are listed in Table 4 , along with rough estimates for the corresponding MEI mutation rates based on the neutral model ( μ\u200a=\u200aθ/ ( 4·Ne ) ) with an effective population size Ne of 10 , 000 56 , 57 ., Confidence intervals for μ and θ ( Table 4 ) take into account Poisson noise and uncertainties in the correction factors , but do not reflect the degree to which the model assumptions are valid ., All three element families have been combined into the allele count spectra shown in Figure 5 , although the Alu family is the dominant component ., Allele frequency spectra for different element families have similar shapes ( Figure 6a ) ., We know from SNP studies that the shape of the allele frequency spectrum is modulated by demographic history , and that this shape is characteristically different for European , African , and Asian populations 56 , 57 ., When compared to SNP allele frequency spectra from the same datasets ( Figure 6b ) , the MEI and SNP frequency spectra show similar trends among the corresponding populations ., Among the three population groups , the CHBJPT spectrum shows relatively few low frequency allele loci ., This was also apparent in comparison with the neutral model ( Figure 5e ) ., We also analyzed population differentiation by applying principal component analysis to the matrix of allele counts across the low coverage pilot samples and loci ( Figures S15 and S16 ) ., Some structure is immediately apparent in the matrix of allele counts , e . g . increased heterozygosity in the YRI samples , but PCA reveals population specific patterns of MEI that result in tight clusters of samples according to geographic origin ( Figure 6c ) ; again similar to population patterns for SNPs 58 , CNVs 59 and deletions 30 ., As few as 39 of the 5 , 370 non-reference MEI loci were located in exonic sequence , mostly in untranslated regions , and only 3 were found in coding exons ( Table 2 ) ., These numbers are much lower than expected from random placement ( Materials and Methods: Functional calculation ) , indicating strong selection against MEI disrupting gene function ., The suppression factor for an MEI to occur in a coding region compared to the genome-averaged rate is 46x , a much stronger suppression than is observed for coding SNPs ( Table S10 , suppression factor\u200a=\u200a3 . 9x ) , and is similar to SNPs that cause the loss of a stop codon ( 42x , derived from Table 2 of 36 ) ., Two of the MEI interrupting coding regions were PCR-validated ., These two MEI appear to be of little functional consequence: ZNF404 is a member of a highly paralogous zinc finger gene family and C14orf166B is a predicted gene without functional annotation ., These findings suggest very strong negative selection against MEI interrupting coding regions ., Although it is obvious from first principles that insertions in functional regions should be deleterious , the observed suppression factor in a large catalog of MEI in populations quantifies the effect ., The high-coverage trio data allows for the most precise estimates of the total number of MEI variants between pairs of individuals because of the high detection sensitivity ., The number of pair-wise variant loci is calculated as the presence or absence of an insertion at a given locus , combining reference and non-reference MEI ., We selected the two trio children ( NA12878 and NA19240 ) for comparison between CEU and YRI individuals and the trio parents for comparison of individuals within the CEU and the YRI population groups ., After corrections for detection sensitivity and false detection ( Text S2 and Table S6 ) , we found 2 , 034±120 MEI variant loci between the African and the European trio children , 1 , 442±120 between the YRI parents , and 663±140 MEI between the CEU parents ., The pair-wise event numbers scale linearly with coalescent time derived from SNPs ( Figure 6d ) in these samples ( Text S2: Coalescent 60–64 ) ., Previous estimates for the de novo mobile element insertion rate and our own estimate of the MEI mutation rate are one event per 20 births in the human population 23 ., Accordingly , we did not expect to find de novo insertions in our sample of two trio children ., Among all MEI events detected in the trio offspring against the reference ( 1 , 778 in NA12878 and 1 , 971 in NA19240 ) , we did identify a single de novo candidate insertion in NA12878 , not detected in either parent or in any other sample ( Table S6 , de Novo ) ., A

Introduction, Results, Discussion, Materials and Methods

As a consequence of the accumulation of insertion events over evolutionary time , mobile elements now comprise nearly half of the human genome ., The Alu , L1 , and SVA mobile element families are still duplicating , generating variation between individual genomes ., Mobile element insertions ( MEI ) have been identified as causes for genetic diseases , including hemophilia , neurofibromatosis , and various cancers ., Here we present a comprehensive map of 7 , 380 MEI polymorphisms from the 1000 Genomes Project whole-genome sequencing data of 185 samples in three major populations detected with two detection methods ., This catalog enables us to systematically study mutation rates , population segregation , genomic distribution , and functional properties of MEI polymorphisms and to compare MEI to SNP variation from the same individuals ., Population allele frequencies of MEI and SNPs are described , broadly , by the same neutral ancestral processes despite vastly different mutation mechanisms and rates , except in coding regions where MEI are virtually absent , presumably due to strong negative selection ., A direct comparison of MEI and SNP diversity levels suggests a differential mobile element insertion rate among populations .

We embarked on this study to explore the 1000 Genomes Project ( 1000GP ) pilot dataset as a substrate for Mobile Element Insertion ( MEI ) discovery and analysis ., MEI is already well known as a significant component of genetic variation in the human population ., However the full extent and effects of MEI can only be assessed by accurate detection in large whole-genome sequencing efforts such as the 1000GP ., In this study we identified 7 , 380 distinct genomic locations of variant MEI and carried out rigorous validation experiments that confirmed the high accuracy of the detected events ., We were able to measure the frequency of each variant in three continental population groups and found that inherited MEI variants propagate through populations in much the same way as single nucleotide polymorphisms , except that MEI are more strongly suppressed in protein coding parts of the genome ., We also found evidence that the MEI mutation rate has not been constant over human population history , rather that different populations appear to have different characteristic MEI mutation rates .

functional genomics, genetic mutation, genome evolution, genome scans, neutral theory, population genetics, genome sequencing, mutation, genome analysis tools, genome databases, mutation types, mutation databases, genetic polymorphism, biology, genetics, genomics, computational biology, genetics and genomics, human genetics

journal.pcbi.1006142

Systematic interrogation of diverse Omic data reveals interpretable, robust, and generalizable transcriptomic features of clinically successful therapeutic targets

More than half of drug candidates that advance beyond phase I clinical trials fail due to lack of efficacy 1 , 2 ., One possible explanation for these failures is sub-optimal target selection 3 ., Many factors must be considered when selecting a target for drug discovery 4 , 5 ., Intrinsic factors include the likelihood of the target to be tractable ( can the target’s activity be altered by a compound , antibody , or other drug modality ? ) , safe ( will altering the target’s activity cause serious adverse events ? ) , and efficacious ( will altering the target’s activity provide significant benefit to patients ? ) ., Extrinsic factors include the availability of investigational reagents and disease models for preclinical target validation , whether biomarkers are known for measuring target engagement or therapeutic effect , the duration and complexity of clinical trials required to prove safety and efficacy , and the unmet need of patients with diseases that might be treated by modulating the target ., Over the past decade , technologies have matured enabling high-throughput genome- , transcriptome- , and proteome-wide profiling of cells and tissues in normal , disease , and experimentally perturbed states ., In parallel , researchers have made substantial progress curating or text-mining biomedical literature to extract and organize information about genes and proteins , such as molecular functions and signaling pathways , into structured datasets ., Taken together , both efforts have given rise to a vast amount of primary , curated , and text-mined data about genes and proteins , which are stored in online repositories and amenable to computational analysis 6 , 7 ., To improve the success rate of drug discovery projects , researchers have investigated whether any features of genes or proteins are useful for target selection ., These computational studies can be categorized according to whether the researchers were trying to predict tractability 8 , 9 , safety 10–13 , efficacy ( no publications to our knowledge ) , or overall success ( alternatively termed “drug target likeness” ) 8 , 13–26 ., Closely related efforts include disease gene prediction , where the goal is to predict genes mechanistically involved in a given disease 27–32 , and disease target prediction , where the goal is to predict genes that would make successful drug targets for a given disease 33–35 ., To our knowledge , we report the first screen for features of genes or proteins that distinguish targets of approved drugs from targets of drug candidates that failed in clinical trials ., In contrast , related prior studies have searched for features that distinguish targets of approved drugs from the rest of the genome ( or a representative subset ) 13 , 15–25 ., Using the remainder of the genome for comparison has been useful for finding features enriched among successful targets , but it is uncertain whether these features are specific to successful targets or are enriched among targets of failed drug candidates as well ., Our study aims to fill this knowledge gap by directly testing for features that separate targets by clinical outcome , expanding the scope of prior studies that have investigated how genetic disease associations 36 and publication trends 37 of targets correlate with clinical outcome ., Our work has five additional innovative characteristics ., First , we included only targets of drugs that are presumed to be selective ( no documented polypharmacology ) to reduce ambiguity in assigning clinical trial outcomes to targets ., Second , we included only phase III failures to enrich for target efficacy failures , as opposed to safety and target engagement failures , which are more common in phase I and phase II 2 ., Third , we excluded targets of assets only indicated for cancer , as studies have observed that features of successful targets for cancer differ from features of successful targets for other indications 22 , 23 , moreover , cancer trials fail more frequently than trials for other indications 2 ., Fourth , we interrogated a diverse and comprehensive set of features , over 150 , 000 features from 67 datasets covering 16 feature types , whereas prior studies have examined only features derived from protein sequence 16–18 , 24 , 25 , protein-protein interactions 13 , 15 , 18–23 , Gene Ontology terms 13 , 15 , 16 , and gene expression profiles 15 , 19 , 21 , 25 ., Fifth , because targets of drugs and drug candidates do not constitute a random sample of the genome , we implemented a suite of tests to assess the robustness and generalizability of features identified as significantly separating successes from failures in the biased sample ., A handful of the initial 150 , 000+ features passed our tests for robustness and generalizability to new targets or target classes ., Interestingly , these features were predominantly derived from gene expression datasets ., Notably , two significant features were discovered repeatedly in multiple datasets: successful targets tended to have lower mean mRNA expression across tissues and higher expression variance than failed targets ., We also trained a classifier to predict phase III success probabilities for untested targets ( no phase III clinical trial outcomes reported for drug candidates that selectively modulate these targets ) ., We identified 943 targets with sufficiently unfavorable expression characteristics to be predicted twice as likely to fail in phase III clinical trials as past phase III targets ., Furthermore , we identified 2 , 700 , 856 target pairs predicted with 99% consistency to have a 2-fold difference in success probability ., Such pairwise comparisons may be useful for prioritizing short lists of targets under consideration for a therapeutic program ., We conclude this paper with a discussion of the biases and limitations faced when attempting to analyze , model , or interpret data on clinical trial outcomes ., We extracted phase III clinical trial outcomes reported in Pharmaprojects 38 for drug candidates reported to be selective ( single documented target ) and tested as treatments for non-cancer diseases ., We grouped the outcomes by target , scored targets with at least one approved drug as successful ( NS = 259 ) , and scored targets with no approved drugs and at least one documented phase III failure as failed ( NF = 72 ) ( S1 Table ) ., The target success rate ( 77% ) appears to be inflated relative to typically reported phase III success rates ( 58% ) 2 because we scored targets by their best outcome across multiple trials ., We obtained target features from the Harmonizome 39 , a recently published collection of features of genes and proteins extracted from over 100 Omics datasets ., We limited our analysis to 67 datasets that are in the public domain or GSK had independently licensed ( Table 1 ) ., Each dataset in the Harmonizome is organized into a matrix with genes labeling the rows and features such as diseases , phenotypes , tissues , and pathways labeling the columns ., We included the mean and standard deviation calculated along the rows of each dataset as additional target features ., These summary statistics provide potentially useful and interpretable information about targets , such as how many pathway associations a target has or how variable a target’s expression is across tissues ., The datasets contained a total of 174 , 228 features covering 16 feature types ( Table 1 ) ., We restricted our analysis to 44 , 092 features that had at least three non-zero values for targets assigned a phase III outcome ., Many datasets had strong correlations among their features ., To reduce feature redundancy and avoid excessive multiple hypothesis testing while maintaining interpretability of features , we replaced each group of highly correlated features with the group mean feature and assigned it a representative label ( Fig 1 , S2 Table ) ., The number of features shrunk to 28 , 562 after reducing redundancy ., We performed permutation tests 40 , 41 on the remaining 28 , 562 target features to find features with a significant difference between the successful and failed targets , and we corrected p-values for multiple hypothesis testing using the Benjamini-Yekutieli method 42 ( Fig 1 , S2 Table ) ., We used permutation testing to apply the same significance testing method to all features , since they had heterogeneous data distributions ., We detected 19 features correlated with clinical outcome at a within-dataset false discovery rate of 0 . 05 ( Table 2 ) ., The significant features were derived from 7 datasets , of which 6 datasets were gene expression atlases: Allen Brain Atlas adult human brain tissues 43 , 44 , Allen Brain Atlas adult mouse brain tissues 43 , 45 , BioGPS human cell types and tissues 46–48 , BioGPS mouse cell types and tissues 46–48 , Genotype-Tissue Expression Project ( GTEx ) human tissues 49 , 50 , and Human Protein Atlas ( HPA ) human tissues 51 ., The remaining dataset , TISSUES 52 , was an integration of experimental gene and protein tissue expression evidence from multiple sources ., Two correlations were significant in multiple datasets: successful targets tended to have lower mean expression across tissues and higher expression variance than failed targets ., Because targets of drugs and drug candidates do not constitute a random sample of the genome , features that separate successful targets from failed targets in our sample may perform poorly as genome-wide predictors of success versus failure ., We performed three analyses to address this issue ( Fig 1 ) ., Statistical significance did not guarantee the remaining features would be useful in practice for discriminating between successes and failures ., To test their utility , we trained a classifier to predict target success or failure , using cross-validation to select a model type ( Random Forest or logistic regression ) and a subset of features useful for prediction ., Because we used all targets with phase III outcomes for the feature selection procedure described above , simply using the final set of features to train a classifier on the same data would yield overly optimistic performance , even with cross-validation ., Therefore , we implemented a nested cross-validation routine to perform both feature selection and model selection 58 ., We searched over 150 , 000 target features from 67 datasets covering 16 feature types for predictors of target success or failure in phase III clinical trials ( Table 1 , Fig 1 ) ., We found several features significantly correlated with phase III outcome , robust to re-sampling , and generalizable across target classes ( Table 2 ) ., To assess the usefulness of such features , we implemented a nested cross-validation routine to select features , train a classifier to predict the probability a target will succeed in phase III clinical trials , and estimate the stability and generalization performance of the model ( Figs 2 and 3 , Tables 3 , 4 and 5 ) ., Ultimately , we found two features useful for predicting success or failure of targets in phase III clinical trials ., Successful targets tended to have low mean mRNA expression across tissues and high standard deviation of mRNA expression across tissues ( Fig 3F ) ., These features were significant in multiple gene expression datasets , which increased our confidence that their relationship to phase III outcome was real , at least for the targets in our sample , which included only targets of selective drugs indicated for non-cancer diseases ., One interpretation of why the gene expression features were predictive of phase III outcome is that they are informative of the specificity of a target’s expression across tissues ., A target with tissue specific expression would have a high standard deviation relative to its mean expression level ., Tissue specific expression has been proposed by us and others as a favorable target characteristic in the past 4 , 14 , 60–62 , but the hypothesis had not been evaluated empirically using examples of targets that have succeeded or failed in clinical trials ., For a given disease , if a target is expressed primarily in the disease tissue , it is considered more likely that a drug will be able to exert a therapeutic effect on the disease tissue while avoiding adverse effects on other tissues ., Additionally , specific expression of a target in the tissue affected by a disease could be an indicator that dysfunction of the target truly causes the disease ., The distribution of the success and failure examples in feature space ( Fig 3F ) partially supports the hypothesis that tissue specific expression is a favorable target feature ., Successes were enriched among targets with low mean expression and high standard deviation of expression ( tissue specific expression ) , and failures were enriched among targets with high mean expression and low standard deviation of expression ( ubiquitous expression ) ., However , it does not hold in general that , at any given mean expression level , targets with high standard deviation of expression tend to be more successful than targets with low standard deviation of expression ., To further investigate the relationship between these features and phase III clinical trial outcomes , we re-ran the entire modeling pipeline ( Fig, 2 ) with gene expression entropy , a feature explicitly quantifying specificity of gene expression across tissues 21 , appended to each tissue expression dataset ( S1 Text ) ., Model performance was unchanged ( S1 Fig ) ; gene expression entropy across tissues became the dominant selected feature , appearing in 610 models over 1000 train-test cycles; and mean gene expression across tissues remained an important feature , appearing in 381 models ( S6 Table ) ., To find concrete examples illustrating when tissue expression may be predictive of clinical trial outcomes , we pulled additional information from the Pharmaprojects database about targets at the two extremes of tissue expression ( tissue specific or ubiquitous ) ., We found examples of:, 1 ) successful tissue specific targets where the target is specifically expressed in the tissue affected by the disease ( Table 6 ) ,, 2 ) failed tissue specific targets with plausible explanations for failure despite tissue specific expression ( Table 7 ) ,, 3 ) failed ubiquitously expressed targets ( Table 8 ) , and, 4 ) successful ubiquitously expressed targets with plausible explanations for success despite ubiquitous expression ( Table 9 ) ., Our results encourage further investigation of the relationship between tissue specific expression and clinical trial outcomes ., Deeper insight may be gleaned from analysis of clinical trial outcomes of target-indication pairs using gene expression features explicitly designed to quantify specificity of a target’s expression in the tissue ( s ) affected by the disease treated in each clinical trial ., Latent factors ( variables unaccounted for in this analysis ) could confound relationships between target features and phase III outcomes ., For example , diseases pursued vary from target to target , and a target’s expression across tissues may be irrelevant for diseases where drugs can be delivered locally or for Mendelian loss-of-function diseases where treatment requires systemic replacement of a missing or defective protein ., Also , clinical trial failure rates vary across disease classes 2 ., Although we excluded targets of cancer therapeutics from our analysis , we otherwise did not control for disease class as a confounding explanatory factor ., Modalities ( e . g . small molecule , antibody , antisense oligonucleotide , gene therapy , or protein replacement ) and directions ( e . g . activation or inhibition ) of target modulation also vary from target to target and could be confounding explanatory factors or alter the dependency between target features and outcomes ., The potential issues described above are symptoms of the fact that our analysis ( and any analysis of clinical trial outcomes ) attempts to draw conclusions from a small ( 331 targets with only 72 failures ) and biased sample 63 , 64 ., The large uncertainty in the performance of the classifier across 200 repetitions of 5-fold cross-validation is evidence of the difficulty in finding robust signal in such a small dataset ( Fig 3 ) ., For example , in the region where the model predicts highest probability of success ( low mean expression and high standard deviation of expression ) , there are no failed phase III targets ( Fig 3F ) , which is why the median PPV rises nearly to 1 ( Fig 3C ) , but targets with phase III outcomes sparsely populate this region , so the PPV varies widely depending upon how targets happen to fall into training and testing sets during cross-validation ., The small sample issue is compounded by latent factors , such as target classes , disease classes , modalities , and directions of target modulation , that are not uniformly represented in the sample ., Correlations between target features and clinical trial outcomes likely depend on these factors , but attempts to stratify , match , or otherwise control for these factors are limited by the sample size ., ( The number of combinations of target class , disease class , modality , and direction of modulation exceeds the sample size . ), We employed several tests to build confidence that our findings generalize across target classes , but did not address other latent factors ., Consequently , we cannot be sure that conclusions drawn from this study apply equally to targets modulated in any direction , by any means , to treat any disease ., For specific cases , expert knowledge and common sense should be relied upon to determine whether conclusions from this study ( or similar studies ) are relevant ., Another limitation is selection bias 63 , 64 ., Targets of drugs are not randomly selected from the genome and cannot be considered representative of the population of all possible targets ., Likewise , diseases treated by drugs are not randomly chosen; therefore , phase III clinical trial outcomes for each target cannot be considered representative of the population of all possible outcomes ., Although we implemented tests to build confidence that our findings can generalize to new targets and new target classes , ultimately , no matter how we dissect the sample , a degree of uncertainty will always remain about the relevance of any findings for new targets that lack a representative counterpart in the sample ., Additionally , data processing and modeling decisions have introduced bias into the analysis ., For example , we restricted the analysis to phase III clinical trial outcomes because failures in phase III are more likely to be due to lack of target efficacy than failures in earlier phases , but factors unrelated to target efficacy still could explain many of the phase III failures , such as poor target engagement , poorly defined clinical trial endpoints , and a poorly defined patient population ., Also , we scored each target as successful or failed by its best outcome in all applicable ( selective drug , non-cancer indication ) phase III clinical trials ., This approach ignores nuances ., A target that succeeded in one trial and failed in all others is treated as equally successful as a target that succeeded in all trials ., Also , the outcome of a target tested in a single trial is treated as equally certain as the outcome of a target tested in multiple trials ., Representing target outcomes as success rates or probabilities may provide better signal for discovering features predictive of outcomes ., Another decision was to use datasets of features as we found them , rather than trying to reason about useful features that could be derived from the original data ., Because of the breadth of data we interrogated , the effort and expertise necessary to hand engineer features equally well across all datasets exceeded our resources ., Others have had success hand engineering features for similar applications in the past , particularly with respect to computing topological properties of targets in protein-protein interaction networks 18 , 20 , 21 ., This analysis could benefit from such efforts , potentially changing a dataset or feature type from yielding no target features correlated with phase III outcomes to yielding one or several useful features 22 ., On a related point , because we placed a priority on discovering interpretable features , we performed dimensionality reduction by averaging groups of highly correlated features and concatenating their ( usually semantically related ) labels ., Dimensionality reduction by principal components analysis 65 or by training a deep auto-encoder 66 could yield more useful features , albeit at the expense of interpretability ., We also employed a stringent univariate feature selection step ( Fig 2 , Step, 2 ) to bias our analysis toward yielding a simple and interpretable model ., In doing so , we diminished the chance of the multivariate feature selection step ( Fig 2 , Step, 4 ) finding highly predictive combinations of features that individually were insignificantly predictive ., We addressed this concern by re-running the entire modeling pipeline ( Fig, 2 ) with the threshold for the univariate feature selection step made less stringent by eliminating the multiple hypothesis testing correction and accepting features with nominal p-values less than 0 . 05 ( S2 Text ) ., This allowed hundreds of features to pass through to the multivariate feature selection step ( Random Forest with incremental feature elimination ) and ultimately dozens of features ( median of 73 ) were selected for each of the final models in the 1000 train-test cycles ( S7 Table ) ., Despite this increase in number of features , the mean expression and standard deviation of expression features were still robustly selected , appearing in 958 and 745 models , respectively , and the models had a median AUROC of 0 . 56 and AUPRC of 0 . 75 , performing no better than the simple models ( S2 Fig ) ., This finding suggests that our sample size was not large enough to robustly select predictive combinations of features from a large pool of candidate features 67 , 68 ., We cannot stress enough the importance of taking care not to draw broad conclusions from our study , particularly with respect to the apparent dearth of features predictive of target success or failure ., We examined only a specific slice of clinical trial outcomes ( phase III trials of selective drugs indicated for non-cancer diseases ) summarized in a particular way ( net outcome per target , as opposed to outcome per target-indication pair ) ., Failure of a feature to be significant in our analysis should not be taken to mean it has no bearing on target selection ., For example , prior studies have quantitatively shown that genetic evidence of disease association ( s ) is a favorable target characteristic 3 , 36 , but we did not find a significant correlation between genetic evidence and target success in phase III clinical trials ., Our finding is consistent with the work of Nelson et al . 36 , who investigated the correlation between genetic evidence and drug development outcomes at all phases and found a significant correlation overall and at all phases of development except phase III ., As a way of checking our work , we applied our methods to test for features that differ between targets of approved drugs and the remainder of the druggable genome ( instead of targets of phase III failures ) , and we recovered the finding of Nelson et al . that targets of approved drugs have significantly more genetic evidence than the remainder of the druggable genome ( S8 Table ) ., This example serves as a reminder to be cognizant of the domain of applicability of research findings ., Though we believe we have performed a rigorous and useful analysis , we have shed light on only a small piece of a large and complex puzzle ., Advances in machine learning enable and embolden us to create potentially powerful predictive models for target selection ., However , as described in the limitations , scarce training data are available , the data are far from ideal , and we must be cautious about building models with biased data and interpreting their predictions ., For example , many features that appeared to be significantly correlated with phase III clinical trial outcomes in our primary analysis did not hold up when we accounted for target class selection bias ., This study highlights the need for both domain knowledge and modeling expertise to tackle such challenging problems ., Our analysis revealed several features that significantly separated targets of approved drugs from targets of drug candidates that failed in phase III clinical trials ., This suggested that it is feasible to construct a model integrating multiple interpretable target features derived from Omics datasets to inform target selection ., Only features derived from tissue expression datasets were promising predictors of success versus failure in phase III , specifically , mean mRNA expression and standard deviation of expression across tissues ., Although these features were significant at a false discovery rate cut-off of 0 . 05 , their effect sizes were too small to be useful for classification of the majority of untested targets , however , even a two-fold improvement in target quality can dramatically increase R&D productivity 69 ., We identified 943 targets predicted to be twice as likely to fail in phase III clinical trials as past phase III targets , and , therefore , should be flagged as having unfavorable expression characteristics ., We also identified 2 , 700 , 856 target pairs predicted with 99% consistency to have a 2-fold difference in success probability , which could be useful for prioritizing short lists of targets with attractive disease relevance ., It should be noted that our analysis was not designed or powered to show that specific datasets or data types have no bearing on target selection ., There are many reasons why a dataset may not have yielded any significant features in our analysis ., In particular , data processing and filtering choices could determine whether or not a dataset or data type has predictive value ., Also , latent factors , such as target classes , disease classes , modalities , and directions of target modulation , could confound or alter the dependency between target features and clinical trial outcomes ., Finally , although we implemented tests to ensure robustness and generalizability of the target features significantly correlated with phase III outcomes , selection bias in the sample of targets available for analysis is a non-negligible limitation of this study and others of its kind ., Nevertheless , we are encouraged by our results and anticipate deeper insights and better models in the future , as researchers improve methods for handling sample biases and learn more informative features ., Our goals in performing dimensionality reduction were to identify groups of highly correlated features , avoid excessive multiple hypothesis testing , and maintain interpretability of features ., For each dataset , we computed pair-wise feature correlations ( r ) using the Spearman correlation coefficient 72–74 for quantitative , filled-in datasets , and the cosine coefficient 73 , 74 for sparse or categorical datasets ., We thresholded the correlation matrix at r2 = 0 . 5 ( for the Spearman correlation coefficient , this corresponds to one feature explaining 50% of the variance of another feature , and for the cosine coefficient , this corresponds to one feature being aligned within 45 degrees of another feature ) and ordered the features by decreasing number of correlated features ., We created a group for the first feature and its correlated features ., If the dataset mean was included in the group , we replaced the group of features with the dataset mean ., Otherwise , we replaced the group of features with the group mean and assigned it the label of the first feature ( to indicate that the feature represents the average of features correlated with the first feature ) , while also retaining a list of the labels of all features included in the group ., We continued through the list of features , repeating the grouping process as described for the first feature , except excluding features already assigned to a group from being assigned to a second group ., We performed permutation tests 40 , 41 to find features with a significant difference between successful and failed targets ., We used permutation testing in order to apply the same significance testing method to all features ., The features in our collection had heterogeneous shapes of their distributions and varying degrees of sparsity , and therefore no single parametric test would be appropriate for all features ., Furthermore , individual features frequently violated assumptions required for parametric tests , such as normality for the t-test ( for continuous-valued features ) or having at least five observations in each entry of the contingency table for the Chi-squared test ( for categorical features ) ., For each feature , we performed 105 success/failure label permutations to obtain a null distribution for the difference between the means of successful and failed targets , and then calculated an empirical two-tailed p-value as the fraction of permutations that yielded a difference between means at least as extreme as the actual observed difference ., We used the Benjamini-Yekutieli method 42 to correct for multiple hypothesis testing within each dataset and accepted features with corrected p-values less than 0 . 05 as significantly correlated with phase III clinical trial outcomes , thus controlling the false discovery rate at 0 . 05 within each dataset ., We trained a classifier to predict target success or failure in phase III clinical trials , using a procedure like the above for initial feature selection , then using cross-validation to select a model type ( Random Forest or logistic regression ) and subset of features useful for prediction ., We used an outer cross-validation loop with 5-folds repeated 200 times , yielding a total of 1000 train-test cycles , to estimate the generalization performance and stability of the feature selection and model selection procedure 58 ., Each train-test cycle had five steps:, 1 ) splitting examples into training and testing sets ,, 2 ) univariate feature selection on the training data ,, 3 ) aggregation of significant features from different datasets into a single feature matrix ,, 4 ) model selection and model-based ( multivariate ) feature selection on the training data , and, 5 ) evaluation of the classifier on the test data ., Computational analyses were written in Python 3 . 4 . 5 and have the following package dependencies: Fastcluster 1 . 1 . 20 , Matplotlib 1 . 5 . 1 , Numpy 1 . 11 . 3 , Requests 2 . 13 . 0 , Scikit-learn 0 . 18 . 1 , Scipy 0 . 18 . 1 , and Statsmodels 0 . 6 . 1 ., Code , documentation , and data have been deposited on GitHub at https://github . com/arouillard/omic-features-successful-targets .

Introduction, Results, Discussion, Conclusion, Methods

Target selection is the first and pivotal step in drug discovery ., An incorrect choice may not manifest itself for many years after hundreds of millions of research dollars have been spent ., We collected a set of 332 targets that succeeded or failed in phase III clinical trials , and explored whether Omic features describing the target genes could predict clinical success ., We obtained features from the recently published comprehensive resource: Harmonizome ., Nineteen features appeared to be significantly correlated with phase III clinical trial outcomes , but only 4 passed validation schemes that used bootstrapping or modified permutation tests to assess feature robustness and generalizability while accounting for target class selection bias ., We also used classifiers to perform multivariate feature selection and found that classifiers with a single feature performed as well in cross-validation as classifiers with more features ( AUROC = 0 . 57 and AUPR = 0 . 81 ) ., The two predominantly selected features were mean mRNA expression across tissues and standard deviation of expression across tissues , where successful targets tended to have lower mean expression and higher expression variance than failed targets ., This finding supports the conventional wisdom that it is favorable for a target to be present in the tissue ( s ) affected by a disease and absent from other tissues ., Overall , our results suggest that it is feasible to construct a model integrating interpretable target features to inform target selection ., We anticipate deeper insights and better models in the future , as researchers can reuse the data we have provided to improve methods for handling sample biases and learn more informative features ., Code , documentation , and data for this study have been deposited on GitHub at https://github . com/arouillard/omic-features-successful-targets .

Drug discovery often begins with a hypothesis that changing the abundance or activity of a target—a biological molecule , usually a protein—will cure a disease or ameliorate its symptoms ., Whether a target hypothesis translates into a successful therapy depends in part on the characteristics of the target , but it is not completely understood which target characteristics are important for success ., We sought to answer this question with a supervised machine learning approach ., We obtained outcomes of target hypotheses tested in clinical trials , scoring targets as successful or failed , and then obtained thousands of features ( i . e . properties or characteristics ) of targets from dozens of biological datasets ., We statistically tested which features differed between successful and failed targets , and built a computational model that used these features to predict success or failure of targets in clinical trials ., We found that successful targets tended to have more variable mRNA abundance from tissue to tissue and lower average abundance across tissues than failed targets ., Thus , it is probably favorable for a target to be present in the tissue ( s ) affected by a disease and absent from other tissues ., Our work demonstrates the feasibility of predicting clinical trial outcomes from target features .

medicine and health sciences, tissue proteins, phase iii clinical investigation, protein expression, clinical medicine, mathematics, artificial intelligence, pharmacology, molecular biology techniques, discrete mathematics, combinatorics, research and analysis methods, computer and information sciences, proteins, gene expression, artificial genetic recombination, gene targeting, molecular biology, molecular biology assays and analysis techniques, gene expression and vector techniques, biochemistry, permutation, drug research and development, phenotypes, clinical trials, genetics, biology and life sciences, physical sciences, machine learning

journal.pcbi.1005389

"Bayesian phylogeography of influenza A/H3N2 for the 2014-15 season in the United States using three(...TRUNCATED)

"Bayesian phylogeography has emerged as a powerful approach to analyzing virus spread ., It utilizes(...TRUNCATED)

Introduction, Results, Discussion

"Ancestral state reconstructions in Bayesian phylogeography of virus pandemics have been improved by(...TRUNCATED)

"For the better part of the last decade , epidemiological researchers have employed a Bayesian frame(...TRUNCATED)

"biogeography, taxonomy, ecology and environmental sciences, medicine and health sciences, pathology(...TRUNCATED)

journal.pgen.1004037

Intrasubtype Reassortments Cause Adaptive Amino Acid Replacements in H3N2 Influenza Genes

"The genome of influenza A virus consists of 8 segments , each represented by an RNA molecule ., Coi(...TRUNCATED)

Introduction, Results, Discussion, Methods

"Reassortments and point mutations are two major contributors to diversity of Influenza A virus; how(...TRUNCATED)

"Influenza A is a rapidly evolving virus with genome composed of eight distinct RNA molecules called(...TRUNCATED)

"organismal evolution, genome evolution, population genetics, microbiology, parallel evolution, muta(...TRUNCATED)

journal.pcbi.1000927

"Synaptic Plasticity Controls Sensory Responses through Frequency-Dependent Gamma Oscillation Resona(...TRUNCATED)

"Synchronous oscillations 1–3 in neural networks are thought to be important to sensory and cognit(...TRUNCATED)

Introduction, Results, Discussion, Methods

"Synchronized gamma frequency oscillations in neural networks are thought to be important to sensory(...TRUNCATED)

"In the nervous system , a network of neurons shows interesting population activities ., One example(...TRUNCATED)

"neuroscience/theoretical neuroscience, biophysics/theory and simulation, computational biology/syst(...TRUNCATED)

journal.pcbi.1002879

An Integrative Model of Ion Regulation in Yeast

"A feature of fungal physiology is the ability to adapt successfully to a variety of environmental p(...TRUNCATED)

Introduction, Results, Discussion, Methods

"Yeast cells are able to tolerate and adapt to a variety of environmental stresses ., An essential a(...TRUNCATED)

"Ion regulation is fundamental to cell physiology ., The concentrations of monovalent ions , such as(...TRUNCATED)

systems biology, biochemical simulations, regulatory networks, biology, computational biology