Spaces:
Running
Running
| import streamlit as st | |
| st.markdown(''' | |
| # Mutual Annotation-Based Prediction of Protein Domain Functions with Domain2GO | |
| ''') | |
| st.markdown( | |
| """ | |
| [](https://www.biorxiv.org/content/10.1101/2022.11.03.514980v1) [](https://github.com/HUBioDataLab/Domain2GO) | |
| """) | |
| st.markdown('<p style="font-size:18px; font-weight:bold">Developers: Erva Ulusoy, Tunca Dogan</p>', unsafe_allow_html=True) | |
| st.markdown( | |
| '<p style="font-size:25px; font-weight:bold">How it works</p>', unsafe_allow_html=True) | |
| st.markdown( | |
| """ | |
| Domain2GO predicts functions of queried proteins by propagating previously generated domain-function association predictions (Domain2GO mapping set). | |
| Domain2GO is developed with the aim of identifying unknown protein functions by associating domains with Gene Ontology terms, thus defining the problem as domain function prediction. Domain2GO mappings are generated using the existing domain and GO annotation data. In order to obtain highly reliable associations, we employed statistical resampling and analyzed the co-occurrence patterns of domains and GO terms on the same proteins. | |
| We applied Domain2GO to predict protein functions, by propagating domain-associated GO terms to proteins that are annotated with those domains. For protein function prediction performance evaluation and comparison against other methods, we employed CAFA3 challenge datasets. The results demonstrated the potential of Domain2GO, especially when predicting molecular function and biological process terms, as it performed better than baseline predictors and curated associations (Fmax = 0.48 and 0.36 for MFO and BPO, respectively). | |
| For more information on the construction of Domain2GO mappings, statistical analysis of mappings, calculation of probability scores and protein function prediction performance evaluation, please refer to our pre-print article: | |
| Ulusoy, E., & Dogan, T. (2022). Mutual Annotation-Based Prediction of Protein Domain Functions with Domain2GO. *bioRxiv*, 514980v1. [Link](https://www.biorxiv.org/content/10.1101/2022.11.03.514980v1) | |
| Overall workflow of Domain2GO is shown below. | |
| """) | |
| st.image('figures/full_methodology.png', width=700) | |
| st.markdown( | |
| '<p style="text-align:center"><em><strong>Schematic representation of the proposed method. (A)</strong> The source datasets were downloaded and organized; <strong>(B)</strong> initial mapping between the InterPro domains and GO terms were obtained, and the mapping parameters were calculated; <strong>(C)</strong> generation of the randomized annotation and mapping sets were constructed; <strong>(D)</strong> co-occurrence similarity distributions were plotted, and thresholds were selected based on statistical resampling; <strong>(E)</strong> an ablation study was conducted by calculating the enrichment of top predictions ranked by different statistical measures and finalized Domain2GO mappings were generated by filtering initial mappings; <strong>(F)</strong> protein function predictions were generated by propagating Domain2GO mappings to target proteins.</em></p>', | |
| unsafe_allow_html=True) | |