--- license: cc-by-nc-sa-4.0 language: - en tags: - biology - genomics pretty_name: Genomics Long Range Benchmark viewer: false --- ## Summary The motivation of the genomics long range benchmark (LRB) is to compile a set of biologically relevant genomic tasks requiring long-range dependencies which will act as a robust evaluation tool for genomic language models. While serving as a strong basis of evaluation, the benchmark must also be efficient and user-friendly. To achieve this we strike a balance between task complexity and computational cost through strategic decisions, such as down-sampling or combining datasets. ## Dataset Tasks The Genomics LRB is a collection of tasks which can be loaded by passing in the corresponding `task_name` into the `load_dataset` function. All of the following datasets allow the user to specify an arbitrarily long sequence length, giving more context to the task, by passing the `sequence_length` kwarg to `load_dataset`. Additional task specific kwargs, if applicable, are mentioned in the sections below.
*Note that as you increase the context length to very large numbers you may start to reduce the size of the dataset since a large context size may cause indexing outside the boundaries of chromosomes. | Task | `task_name` | Sample Output | # Train Seqs | # Test Seqs | |--------------------------------------|----------------------------------------|-------------------------------------------------------------------------------------------| ------------ | ----------- | | Variant Effect Causal eQTL | `variant_effect_causal_eqtl` | {ref sequence, alt sequence, label, tissue, chromosome,position, distance to nearest TSS} | 88717 | 8846 | | Variant Effect Pathogenic Clinvar | `variant_effect_pathogenic_clinvar` | {ref sequence, alt sequence, label, chromosome, position} | 38634 | 1018 | | Variant Effect Pathogenic OMIM | `variant_effect_pathogenic_omim` | {ref sequence, alt sequence, label,chromosome, position} | - | 2321473 | | CAGE Prediction | `cage_prediction` | {sequence, labels, chromosome} | 33891 | 1922 | | Bulk RNA Expression | `bulk_rna_expression` | {sequence, labels, chromosome,position} | 22827 | 990 | | Chromatin Features Histone_Marks | `chromatin_features_histone_marks` | {sequence, labels,chromosome, position} | 2203689 | 227456 | | Chromatin Features DNA_Accessibility | `chromatin_features_dna_accessibility` | {sequence, labels,chromosome, position} | 2203689 | 227456 | | Regulatory Elements Promoter | `regulatory_element_promoter` | {sequence, label,chromosome, start, stop} | 953376 | 96240 | | Regulatory Elements Enhancer | `regulatory_element_enhancer` | {sequence, label,chromosome, start, stop} | 1914575 | 192201 | ## Usage Example ```python from datasets import load_dataset # Use this parameter to download sequences of arbitrary length (see docs below for edge cases) sequence_length=2048 # One of: # ["variant_effect_causal_eqtl","variant_effect_pathogenic_clinvar", # "variant_effect_pathogenic_omim","cage_prediction", "bulk_rna_expression", # "chromatin_features_histone_marks","chromatin_features_dna_accessibility", # "regulatory_element_promoter","regulatory_element_enhancer"] task_name = "variant_effect_causal_eqtl" dataset = load_dataset( "InstaDeepAI/genomics-long-range-benchmark", task_name=task_name, sequence_length=sequence_length, ) ``` ### 1. CAGE Prediction Cap Analysis Gene Expression(CAGE) is a biological assay used to measure the level of mRNA production rather than steady state values, taking into account both production and degradation. Being able to accurately predict mRNA levels as measured by CAGE is essential for deciphering tissue-specific expression patterns, transcriptional networks, and identifying differentially expressed genes with functional significance. #### Source Original CAGE data comes from FANTOM5. We used processed labeled data obtained from the [Basenji paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5932613/) which also used to train Enformer and is located [here](https://console.cloud.google.com/storage/browser/basenji_barnyard/data/human?pageState=(%22StorageObjectListTable%22:(%22f%22:%22%255B%255D%22))&prefix=&forceOnObjectsSortingFiltering=false). Sequence data originates from the GRCh38 genome assembly. #### Data Processing The original dataset from the Basenji paper includes labels for 638 CAGE total tracks over 896 bins (each bin corresponding to 128 base pairs) totaling over ~70 GB. In the interest of dataset size and user-friendliness, only a subset of the labels are selected. From the 638 CAGE tracks, 50 of these tracks are selected with the following criteria: 1. Only select one cell line 2. Only keep mock treated and remove other treatments 3. Only select one donor The [896 bins, 50 tracks] labels total in at ~7 GB. A description of the 50 included CAGE tracks can be found here `cage_prediction/label_mapping.csv`. #### Task Structure Type: Multi-variable regression
Because this task involves predicting expression levels for 128bp bins and there are 896 total bins in the dataset, there are in essence labels for 896 * 128 = 114,688 basepair sequences. If you request a sequence length smaller than 114,688 bps than the labels will be subsetted. Task Args:
`sequence_length`: an interger type, the desired final sequence length, *must be a multiple of 128 given the binned nature of labels
Input: a genomic nucleotide sequence
Output: a variable length vector depending on the requested sequence length [requested_sequence_length / 128, 50] #### Splits Train/Test splits were maintained from Basenji and Enformer where randomly sampling was used to generate the splits. Note that for this dataset a validation set is also returned. In practice we merged the validation set with the train set and use cross validation to select a new train and validation set from this combined set. --- ### 2. Bulk RNA Expression In comparison to CAGE, bulk RNA sequencing assays measure the steady state level (both transription and degradation) of mRNA in a population of cells. #### Source Original data comes from GTEx. We use processed data files from the [ExPecto paper](https://www.nature.com/articles/s41588-018-0160-6) found [here](https://github.com/FunctionLab/ExPecto/tree/master/resources). Sequence data originates from the GRCh37/hg19 genome assembly. #### Data Processing The continuous labels were log(1+x) transformed and standardized. A list of names of tissues corresponding to the labels can be found here: `bulk_rna_expression/label_mapping.csv`. #### Task Structure Type: Multi-variable regression
Task Args:
`sequence_length`: an interger type, the desired final sequence length
Input: a genomic nucleotide sequence centered around the CAGE representative trancription start site
Output: a 218 length vector of continuous values corresponding to the bulk RNA expression levels in 218 different tissue types #### Splits Train: chromosomes 1-7,9-22,X,Y
Test: chromosome 8 #### Metrics Mean Spearman correlation across tissues
Mean Spearman correlation across genes
R2 --- ### 3. Variant Effect Causal eQTL In genomics, a key objective is to predict how genetic variants affect gene expression in specific cell types. #### Source Original data comes from GTEx. However, we used processed data files from the [Enformer paper](https://www.nature.com/articles/s41592-021-01252-x) located [here](https://console.cloud.google.com/storage/browser/dm-enformer/data/gtex_fine/vcf?pageState=(%22StorageObjectListTable%22:(%22f%22:%22%255B%255D%22))&prefix=&forceOnObjectsSortingFiltering=false). Sequence data originates from the GRCh38 genome assembly. #### Data Processing In Enformer the datasets were partitioned in 48 different sets based on the tissue types. In our framing of the task we combine all samples across all tissues into one set and provide the tissue type along with each sample. As data files were used from Enformer, the labels were constructed according to their methodology - variants were labeled as 1 if their posterior inclusion probability was greater than 0.9 as assigned by the population-based fine-mapping tool SuSiE, while a matched set of negative variants was built with posterior inclusion probabilities of less than .01. #### Task Structure Type: Binary classification
Task Args:
`sequence_length`: an interger type, the desired final sequence length
Input: a genomic nucleotide sequence centered on the SNP with the reference allele at the SNP location, a genomic nucleotide sequence centered on the SNP with the alternative allele at the SNP location, and tissue type
Output: a binary value refering to whether the variant has an effect on gene expression #### Splits Train: chromosomes 1-8, 11-22, X, Y
Test: chromosomes 9,10 --- ### 4. Variant Effect Pathogenic ClinVar This task differs from the above in that variants in this datasets are linked to disease pathogenicity rather than effect on gene expression. Specifically, this tasks involves predicting whether a variant located in a coding region is pathogenic or is a common missense variant. #### Source Original data comes from ClinVar and gnomAD. However, we use processed data files from the [GPN-MSA paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10592768/) located here :https://huggingface.co/datasets/songlab/human_variants/blob/main/test.parquet . Sequence data originates from the GRCh38 genome assembly. #### Data Processing Human labelers generated the pathogenic missense set from ClinVar. The non-pathogenic set comes from common missense variants in gnomAD. #### Task Structure Type: Binary classification
Task Args:
`sequence_length`: an interger type, the desired final sequence length
Input: a genomic nucleotide sequence centered on the SNP with the reference allele at the SNP location, a genomic nucleotide sequence centered on the SNP with the alternative allele at the SNP location
Output: a binary value referring to whether the variant is pathogenic or not #### Splits Train: chromosomes 1-7, 9-22, X, Y
Test: chromosomes 8 --- ### 5. Variant Effect Pathogenic OMIM This tasks involves predicting whether a variant located in a regulatory region is implicated in a mendelian disease or is a common missense variant. #### Source Original data comes from the Online Mendelian Inheritance in Man (OMIM) and gnomAD databases. However, we use processed data files from the [GPN-MSA paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10592768/) located here :https://huggingface.co/datasets/songlab/omim/blob/main/test.parquet. Sequence data originates from the GRCh38 genome assembly. #### Data Processing Pathogenic variants originate from OMIM. The non-pathogenic set comes from common missense variants in gnomAD. #### Task Structure Type: Binary classification
Task Args:
`sequence_length`: an interger type, the desired final sequence length
`subset`: a boolean type, whether to use the full dataset or a subset of the dataset (we provide this option as the full dataset has millions of samples) Input: a genomic nucleotide sequence centered on the SNP with the reference allele at the SNP location, a genomic nucleotide sequence centered on the SNP with the alternative allele at the SNP location
Output: a binary value referring to whether the variant is pathogenic or not #### Splits Test: all chromosomes --- ### 6. Chromatin Features Chromatin refers to the mixture of DNA and proteins that form the chromosomes and allow the DNA to exist in a compact structure. There are many factors describing the state of chromatin which play important roles in regulating the cell. In this task we present two features which represent chromatin state, histone marks and DNA accessibility. Histones are proteins which in combination with a short strand of DNA form the basic unit of chromatin called the nucleosome. Histone marks correspond to the chemical modifications of histones which include methylation, acetylation, phosphorylation, and ubiquitination. DNA accessibility as measured through DNase I–hypersensitive sites refers to regions of the chromosome where DNA is not bound in a nucleosome but rather is open and accessible for other proteins such as transcription factors to bind. #### Source Original data used to generate labels for histone marks and DNase profiles comes from the ENCODE and Roadmap Epigenomics project. We used processed data files from the [Deep Sea paper](https://www.nature.com/articles/nmeth.3547) to build this dataset. Sequence data originates from the GRCh37/hg19 genome assembly. #### Data Processing As we use the data from Deep Sea the labels were generated according to the their following procedure - the genome was split into 200 base pairs bins and bins were labeled as 1 if more than half of the bin overlapped with the peak region of the chromatin feature annotations and 0 otherwise. Data from Deep Sea contains 919 labels for various chromatin features - 690 TF binding profiles, 125 DNase profiles and 104 histone-mark profiles. To make the dataset more accessible while still providing a representative set, we randomly sampled 20 histone marks and 20 DNase tissues. #### Task Structure Type: Multi-label binary classification Task Args:
`sequence_length`: an interger type, the desired final sequence length
`subset`: a boolean type, whether to use the full dataset or a subset of the dataset (we provide this option as the full dataset has millions of samples) Input: a genomic nucleotide sequence centered on the 200 base pair bin that is associated with the labels
Output: a vector of length 20 with binary entries #### Splits Train set: chromosomes 1-7,10-22
Test set: chromosomes 8,9 --- ### 7. Regulatory Element Regulatory elements are segments of DNA that play a pivotal role in controlling gene expression and cellular processes. These elements include promoters, enhancers, and other regions that interact with transcription factors and other proteins to modulate the activity of genes. Identifying regulatory elements is essential for understanding the complex regulatory networks governing cellular functions. #### Source Original data annotations to build labels came from the Search Candidate cis-Regulatory Elements by ENCODE project. Sequence data originates from the GRCh38 genome assembly. #### Data Processing To generate labels from the annotations we first split the genome into 200 base pairs bins. We label a bin as 1 if more than half of the bin overlaps with any regulatory element given in the annotations and 0 otherwise. #### Task Structure Type: Binary classification Task Args:
`sequence_length`: an interger type, the desired final sequence length
`subset`: a boolean type, whether to use the full dataset or a subset of the dataset (we provide this option as the full dataset has millions of samples) Input: a genomic nucleotide sequence centered on the 200 base pair bin that is associated with the label
Output: a single binary value #### Splits Train set: chromosomes 1-7,10-22
Test set: chromosomes 8,9