etrop commited on
Commit
84cfc69
1 Parent(s): da2e1f1

add _new_datasets (#2)

Browse files

- git lfs track new datasets (6f33aef1b1084c2f366ab6fcde05c403511f1ab5)
- add new datasets (e5de966305a3236ebc83263aa73141d98a23209e)
- refactor vep causal eqtl name (fb5e83ea8c8da8784c60c3466547b6947f473443)
- refactor folder name (0e903895e3f050c79d2d7ac3844a2d7e17781219)
- update files to be lfs tracked (98eadaf1a715eed8381c32f3c30e9a5d6899c7f8)
- update main loader script for new datasets (ca412e99e53c186014c69a30e05756f35c4d4024)
- fix file path for causal eqtl data (561af90663da0da6aaee642622e0b661744e742a)
- fix file names for pathogenic datasets (cc38ba3856a08d6401eeff1a36fe5b8dafc1cfe1)
- fix label naming chromatin features (461969927800e16441b1362f60613ed4cb74f12f)
- fix cage file path naming (07eece405aaeb544ff18813d84f9374152352ea8)
- update loading script (1575e22e185371f31365424ed36ba54a547910ff)
- update readme for additional datasets (46d85ff2b9ed6819b26d06fc35d9651d60e51abc)
- update README (66099dcefd0a4c704f334718dafb33c23fefbf74)
- update the return values for cage and regulatory elements (634eabc0c55af9cfa443b4caea908949a20514e3)
- update readme for return values (2f79d04a93ddddeae89b3ad72c9f6c47fc8b5549)
- update the return values for cage (4504d2df698616dd2ed3f3300b7925e6c3036152)
- missing punctuation (ea6154d266f23abce923a09961af9cdc2164e41b)

.gitattributes CHANGED
@@ -54,3 +54,13 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
54
  *.jpeg filter=lfs diff=lfs merge=lfs -text
55
  *.webp filter=lfs diff=lfs merge=lfs -text
56
  rna_expression_values.csv filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
54
  *.jpeg filter=lfs diff=lfs merge=lfs -text
55
  *.webp filter=lfs diff=lfs merge=lfs -text
56
  rna_expression_values.csv filter=lfs diff=lfs merge=lfs -text
57
+ chromatin_features/histones_and_dnase_subset.csv filter=lfs diff=lfs merge=lfs -text
58
+ chromatin_features/histones_and_dnase.csv filter=lfs diff=lfs merge=lfs -text
59
+ regulatory_elements/enhancer_dataset.csv filter=lfs diff=lfs merge=lfs -text
60
+ regulatory_elements/enhancer_dataset_subset.csv filter=lfs diff=lfs merge=lfs -text
61
+ regulatory_elements/promoter_dataset.csv filter=lfs diff=lfs merge=lfs -text
62
+ regulatory_elements/promoter_dataset_subset.csv filter=lfs diff=lfs merge=lfs -text
63
+ variant_effect_pathogenic/vep_pathogenic_coding.csv filter=lfs diff=lfs merge=lfs -text
64
+ variant_effect_pathogenic/vep_pathogenic_non_coding.csv filter=lfs diff=lfs merge=lfs -text
65
+ variant_effect_pathogenic/vep_pathogenic_non_coding_subset.csv filter=lfs diff=lfs merge=lfs -text
66
+ variant_effect_causal_eqtl/All_Tissues.csv filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -3,30 +3,40 @@ license: cc-by-nc-sa-4.0
3
  language:
4
  - en
5
  tags:
6
- - biology
7
- - genomics
8
- pretty_name: Genomics Long Range Benchmark
 
 
9
  viewer: false
10
  ---
11
 
12
  ## Summary
13
- 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.
 
14
  While serving as a strong basis of evaluation, the benchmark must also be efficient and user-friendly.
15
  To achieve this we strike a balance between task complexity and computational cost through strategic decisions, such as down-sampling or combining datasets.
16
 
17
- ## Dataset Tasks
18
- 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
19
- allow the user to specify an arbitrarily long sequence length, giving more context to the task, by passing `sequence_length` kwarg to `load_dataset`. Additional task specific kwargs, if applicable,
20
- are mentioend in the sections below.<br>
 
 
21
  *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
22
  cause indexing outside the boundaries of chromosomes.
23
 
24
- | Task | `task_name` | Sample Output | # Train Seqs | # Test Seqs |
25
- | --------- | ---------- | ------ | ------------ | ----------- |
26
- | CAGE Prediction | `cage_prediction`| {sequence, labels, chromosome} | 36086 | 1922 |
27
- | Bulk RNA Expression | `bulk_rna_expression` | {sequence, labels, chromosome} | 22827 | 990 |
28
- | Variant Effect Gene Expression | `variant_effect_gene_expression` | {ref sequence, alt sequence, label, tissue, chromosome, distance to nearest TSS} | 89060 | 8862 |
29
-
 
 
 
 
 
30
 
31
  ## Usage Example
32
  ```python
@@ -35,8 +45,13 @@ from datasets import load_dataset
35
  # Use this parameter to download sequences of arbitrary length (see docs below for edge cases)
36
  sequence_length=2048
37
 
38
- # One of ["cage_prediction", "bulk_rna_expression", "variant_effect_gene_expression"]
39
- task_name = "variant_effect_gene_expression"
 
 
 
 
 
40
 
41
  dataset = load_dataset(
42
  "InstaDeepAI/genomics-long-range-benchmark",
@@ -46,20 +61,128 @@ dataset = load_dataset(
46
 
47
  ```
48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
49
 
 
50
 
51
- ### 1. CAGE Prediction
52
- 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
53
- degradation. Being able to accurately predict mRNA levels as measured by CAGE is essential for deciphering tissue-specific expression patterns, transcriptional networks, and
54
- identifying differentially expressed genes with functional significance.
 
 
 
 
 
 
 
 
 
 
 
 
 
55
 
56
  #### Source
57
- 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).
 
 
58
  Sequence data originates from the GRCh38 genome assembly.
59
 
60
  #### Data Processing
61
  The original dataset from the Basenji paper includes labels for 638 CAGE total tracks over 896 bins (each bin corresponding to 128 base pairs)
62
- totaling over ~70 GB. In the interest of dataset size and user friendliness, only a subset of the labels are selected.
 
63
  From the 638 CAGE tracks, 50 of these tracks are selected with the following criteria:
64
 
65
  1. Only select one cell line
@@ -67,6 +190,7 @@ From the 638 CAGE tracks, 50 of these tracks are selected with the following cri
67
  3. Only select one donor
68
 
69
  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`.
 
70
 
71
  #### Task Structure
72
 
@@ -75,7 +199,7 @@ Because this task involves predicting expression levels for 128bp bins and there
75
  you request a sequence length smaller than 114,688 bps than the labels will be subsetted.
76
 
77
  Task Args:<br>
78
- `sequence_length`: an interger type, the desired final sequence length, *must be a multiple of 128 given the binned nature of labels<br>
79
 
80
  Input: a genomic nucleotide sequence<br>
81
  Output: a variable length vector depending on the requested sequence length [requested_sequence_length / 128, 50]
@@ -84,30 +208,36 @@ Output: a variable length vector depending on the requested sequence length [req
84
  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
85
  set with the train set and use cross validation to select a new train and validation set from this combined set.
86
 
87
- #### Metrics
88
- Mean Pearson correlation across tracks - compute Pearson correlation for a track using all positions for all genes in the test set, then mean over all tracks <br>
89
- Mean Pearson correlation across genes - compute Pearson correlation for a gene using all positions and all tracks, then mean over all genes in the test set <br>
90
- R<sup>2</sup>
91
-
92
 
93
  ---
94
 
95
- ### 2. Bulk RNA Expression
96
- In comparison to CAGE, bulk RNA sequencing assays measure the steady state level (both transription and degradation) of mRNA in a population of cells.
 
 
97
 
98
  #### Source
99
  Original data comes from GTEx. We use processed data files from the [ExPecto paper](https://www.nature.com/articles/s41588-018-0160-6) found
100
  [here](https://github.com/FunctionLab/ExPecto/tree/master/resources). Sequence data originates from the GRCh37/hg19 genome assembly.
101
 
102
  #### Data Processing
103
- 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`.
 
 
 
 
 
 
 
 
 
104
 
105
  #### Task Structure
106
 
107
  Type: Multi-variable regression<br>
108
 
109
  Task Args:<br>
110
- `sequence_length`: an interger type, the desired final sequence length<br>
111
 
112
  Input: a genomic nucleotide sequence centered around the CAGE representative trancription start site<br>
113
  Output: a 218 length vector of continuous values corresponding to the bulk RNA expression levels in 218 different tissue types
@@ -116,42 +246,70 @@ Output: a 218 length vector of continuous values corresponding to the bulk RNA e
116
  Train: chromosomes 1-7,9-22,X,Y<br>
117
  Test: chromosome 8
118
 
119
- #### Metrics
120
- Mean Spearman correlation across tissues <br>
121
- Mean Spearman correlation across genes <br>
122
- R<sup>2</sup>
123
-
124
  ---
125
-
126
- ### 3. Variant Effect Gene Expression
127
- In genomics, a key objective is to predict how genetic variants affect gene expression in specific cell types.
128
 
129
  #### Source
130
- 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).
131
- Sequence data originates from the GRCh38 genome assembly.
132
 
133
  #### Data Processing
134
- 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
135
- and provide the tissue type along with each sample.
136
-
137
- 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
138
- 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.
 
 
 
 
139
 
140
  #### Task Structure
141
 
142
- Type: Binary classification<br>
143
 
144
  Task Args:<br>
145
- `sequence_length`: an interger type, the desired final sequence length<br>
 
146
 
147
- 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<br>
148
- Output: a binary value refering to whether the variant has an effect on gene expression
149
 
150
  #### Splits
151
- Train: chromosomes 1-8, 11-22, X, Y<br>
152
- Test: chromosomes 9,10
 
 
 
 
 
 
 
 
 
153
 
154
- #### Metrics
155
- Accuracy<br>
156
- AUROC<br>
157
- AUPRC
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  language:
4
  - en
5
  tags:
6
+ - Genomics
7
+ - Benchmarks
8
+ - Language Models
9
+ - DNA
10
+ pretty_name: Genomics Long-Range Benchmark
11
  viewer: false
12
  ---
13
 
14
  ## Summary
15
+ The motivation of the genomics long-range benchmark (LRB) is to compile a set of
16
+ biologically relevant genomic tasks requiring long-range dependencies which will act as a robust evaluation tool for genomic language models.
17
  While serving as a strong basis of evaluation, the benchmark must also be efficient and user-friendly.
18
  To achieve this we strike a balance between task complexity and computational cost through strategic decisions, such as down-sampling or combining datasets.
19
 
20
+ ## Benchmark Tasks
21
+ The Genomics LRB is a collection of nine tasks which can be loaded by passing in the
22
+ corresponding `task_name` into the `load_dataset` function. All of the following datasets
23
+ allow the user to specify an arbitrarily long sequence length, giving more context
24
+ to the task, by passing the `sequence_length` kwarg to `load_dataset`. Additional task
25
+ specific kwargs, if applicable, are mentioned in the sections below.<br>
26
  *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
27
  cause indexing outside the boundaries of chromosomes.
28
 
29
+ | Task | `task_name` | Sample Output | ML Task Type | # Outputs | # Train Seqs | # Test Seqs | Data Source |
30
+ |-------|-------------|-------------------------------------------------------------------------------------------|-------------------------|-------------|--------------|----------- |----------- |
31
+ | Variant Effect Causal eQTL | `variant_effect_causal_eqtl` | {ref sequence, alt sequence, label, tissue, chromosome,position, distance to nearest TSS} | SNP Classification | 1 | 88717 | 8846 | GTEx (via [Enformer](https://www.nature.com/articles/s41592-021-01252-x)) |
32
+ | Variant Effect Pathogenic ClinVar | `variant_effect_pathogenic_clinvar` | {ref sequence, alt sequence, label, chromosome, position} | SNP Classification | 1 | 38634 | 1018 | ClinVar, gnomAD (via [GPN-MSA](https://www.biorxiv.org/content/10.1101/2023.10.10.561776v1)) |
33
+ | Variant Effect Pathogenic OMIM | `variant_effect_pathogenic_omim` | {ref sequence, alt sequence, label,chromosome, position} | SNP Classification | 1 | - | 2321473 |OMIM, gnomAD (via [GPN-MSA](https://www.biorxiv.org/content/10.1101/2023.10.10.561776v1)) |
34
+ | CAGE Prediction | `cage_prediction` | {sequence, labels, chromosome,label_start_position,label_stop_position} | Binned Regression | 50 per bin | 33891 | 1922 | FANTOM5 (via [Basenji](https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1008050)) |
35
+ | Bulk RNA Expression | `bulk_rna_expression` | {sequence, labels, chromosome,position} | Seq-wise Regression | 218 | 22827 | 990 | GTEx, FANTOM5 (via [ExPecto](https://www.nature.com/articles/s41588-018-0160-6)) |
36
+ | Chromatin Features Histone_Marks | `chromatin_features_histone_marks` | {sequence, labels,chromosome, position, label_start_position,label_stop_position} | Seq-wise Classification | 20 | 2203689 | 227456 | ENCODE, Roadmap Epigenomics (via [DeepSea](https://pubmed.ncbi.nlm.nih.gov/30013180/) |
37
+ | Chromatin Features DNA_Accessibility | `chromatin_features_dna_accessibility` | {sequence, labels,chromosome, position, label_start_position,label_stop_position} | Seq-wise Classification | 20 | 2203689 | 227456 | ENCODE, Roadmap Epigenomics (via [DeepSea](https://pubmed.ncbi.nlm.nih.gov/30013180/)) |
38
+ | Regulatory Elements Promoter | `regulatory_element_promoter` | {sequence, label,chromosome, start, stop, label_start_position,label_stop_position} | Seq-wise Classification | 1| 953376 | 96240 | SCREEN |
39
+ | Regulatory Elements Enhancer | `regulatory_element_enhancer` | {sequence, label,chromosome, start, stop, label_start_position,label_stop_position} | Seq-wise Classification | 1| 1914575 | 192201 | SCREEN |
40
 
41
  ## Usage Example
42
  ```python
 
45
  # Use this parameter to download sequences of arbitrary length (see docs below for edge cases)
46
  sequence_length=2048
47
 
48
+ # One of:
49
+ # ["variant_effect_causal_eqtl","variant_effect_pathogenic_clinvar",
50
+ # "variant_effect_pathogenic_omim","cage_prediction", "bulk_rna_expression",
51
+ # "chromatin_features_histone_marks","chromatin_features_dna_accessibility",
52
+ # "regulatory_element_promoter","regulatory_element_enhancer"]
53
+
54
+ task_name = "variant_effect_causal_eqtl"
55
 
56
  dataset = load_dataset(
57
  "InstaDeepAI/genomics-long-range-benchmark",
 
61
 
62
  ```
63
 
64
+ ### 1. Variant Effect Causal eQTL
65
+ Predicting the effects of genetic variants, particularly expression quantitative trait loci (eQTLs), is essential for understanding the molecular basis of several diseases.
66
+ eQTLs are genomic loci that are associated with variations in mRNA expression levels among individuals.
67
+ By linking genetic variants to causal changes in mRNA expression, researchers can
68
+ uncover how certain variants contribute to disease development.
69
+
70
+ #### Source
71
+ Original data comes from GTEx. Processed data in the form of vcf files for positive
72
+ and negative variants across 49 different tissue types were obtained from the
73
+ [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=%28%22StorageObjectListTable%22:%28%22f%22:%22%255B%255D%22%29%29&prefix=&forceOnObjectsSortingFiltering=false).
74
+ Sequence data originates from the GRCh38 genome assembly.
75
+
76
+ #### Data Processing
77
+ Fine-mapped GTEx eQTLs originate from [Wang et al](https://www.nature.com/articles/s41467-021-23134-8), while the negative matched set of
78
+ variants comes from [Avsec et al](https://www.nature.com/articles/s41592-021-01252-x)
79
+ . The statistical fine-mapping tool SuSiE was used to label variants.
80
+ Variants from the fine-mapped eQTL set were selected and given positive labels if
81
+ their posterior inclusion probability was > 0.9,
82
+ as assigned by SuSiE. Variants from the matched negative set were given negative labels if their
83
+ posterior inclusion probability was < 0.01.
84
+
85
+ #### Task Structure
86
+
87
+ Type: Binary classification<br>
88
+
89
+ Task Args:<br>
90
+ `sequence_length`: an integer type, the desired final sequence length<br>
91
+
92
+ 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<br>
93
+ Output: a binary value referring to whether the variant has a causal effect on gene
94
+ expression
95
+
96
+ #### Splits
97
+ Train: chromosomes 1-8, 11-22, X, Y<br>
98
+ Test: chromosomes 9,10
99
+
100
+ ---
101
+
102
+ ### 2. Variant Effect Pathogenic ClinVar
103
+ A coding variant refers to a genetic alteration that occurs within the protein-coding regions of the genome, also known as exons.
104
+ Such alterations can impact protein structure, function, stability, and interactions
105
+ with other molecules, ultimately influencing cellular processes and potentially contributing to the development of genetic diseases.
106
+ Predicting variant pathogenicity is crucial for guiding research into disease mechanisms and personalized treatment strategies, enhancing our ability to understand and manage genetic disorders effectively.
107
+
108
+ #### Source
109
+ Original data comes from ClinVar and gnomAD. However, we use processed data files
110
+ from the [GPN-MSA paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10592768/)
111
+ located [here](https://huggingface.co/datasets/songlab/human_variants/blob/main/test.parquet).
112
+ Sequence data originates from the GRCh38 genome assembly.
113
+
114
+ #### Data Processing
115
+ Positive labels correspond to pathogenic variants originating from ClinVar whose review status was
116
+ described as having at least a single submitted record with a classification but without assertion criteria.
117
+ The negative set are variants that are defined as common from gnomAD. gnomAD version 3.1.2 was downloaded and filtered to variants with allele number of at least 25,000. Common
118
+ variants were defined as those with MAF > 5%.
119
+
120
+ #### Task Structure
121
+
122
+ Type: Binary classification<br>
123
+
124
+ Task Args:<br>
125
+ `sequence_length`: an integer type, the desired final sequence length<br>
126
+
127
+ 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<br>
128
+ Output: a binary value referring to whether the variant is pathogenic or not
129
+
130
+ #### Splits
131
+ Train: chromosomes 1-7, 9-22, X, Y<br>
132
+ Test: chromosomes 8
133
+
134
+ ---
135
+
136
+ ### 3. Variant Effect Pathogenic OMIM
137
+ Predicting the effects of regulatory variants on pathogenicity is crucial for understanding disease mechanisms.
138
+ Elements that regulate gene expression are often located in non-coding regions, and variants in these areas can disrupt normal cellular function, leading to disease.
139
+ Accurate predictions can identify biomarkers and therapeutic targets, enhancing personalized medicine and genetic risk assessment.
140
+
141
+ #### Source
142
+ Original data comes from the Online Mendelian Inheritance in Man (OMIM) and gnomAD
143
+ databases.
144
+ However, we use processed data files from the
145
+ [GPN-MSA paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10592768/) located [here](
146
+ https://huggingface.co/datasets/songlab/omim/blob/main/test.parquet).
147
+ Sequence data originates from the GRCh38 genome assembly.
148
+
149
+ #### Data Processing
150
+ Positive labeled data originates from a curated set of pathogenic variants located
151
+ in the Online Mendelian Inheritance in Man (OMIM) catalog. The negative set is
152
+ composed of variants that are defined as common from gnomAD. gnomAD version 3.1.2 was downloaded and filtered to variants with
153
+ allele number of at least 25,000. Common variants were defined as those with minor allele frequency
154
+ (MAF) > 5%.
155
 
156
+ #### Task Structure
157
 
158
+ Type: Binary classification<br>
159
+
160
+ Task Args:<br>
161
+ `sequence_length`: an integer type, the desired final sequence length<br>
162
+ `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)
163
+
164
+ 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<br>
165
+ Output: a binary value referring to whether the variant is pathogenic or not
166
+
167
+ #### Splits
168
+ Test: all chromosomes
169
+
170
+ ---
171
+
172
+ ### 4. CAGE Prediction
173
+ CAGE provides accurate high-throughput measurements of RNA expression by mapping TSSs at a nucleotide-level resolution.
174
+ This is vital for detailed mapping of TSSs, understanding gene regulation mechanisms, and obtaining quantitative expression data to study gene activity comprehensively.
175
 
176
  #### Source
177
+ Original CAGE data comes from FANTOM5. We used processed labeled data obtained from
178
+ the [Basenji paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5932613/) which
179
+ also used to train Enformer and is located [here](https://console.cloud.google.com/storage/browser/basenji_barnyard/data/human?pageState=%28%22StorageObjectListTable%22:%28%22f%22:%22%255B%255D%22%29%29&prefix=&forceOnObjectsSortingFiltering=false).
180
  Sequence data originates from the GRCh38 genome assembly.
181
 
182
  #### Data Processing
183
  The original dataset from the Basenji paper includes labels for 638 CAGE total tracks over 896 bins (each bin corresponding to 128 base pairs)
184
+ totaling over ~70 GB. In the interest of dataset size and user-friendliness, only a
185
+ subset of the labels are selected.
186
  From the 638 CAGE tracks, 50 of these tracks are selected with the following criteria:
187
 
188
  1. Only select one cell line
 
190
  3. Only select one donor
191
 
192
  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`.
193
+ *Note the data in this repository for this task has not already been log(1+x) normalized.
194
 
195
  #### Task Structure
196
 
 
199
  you request a sequence length smaller than 114,688 bps than the labels will be subsetted.
200
 
201
  Task Args:<br>
202
+ `sequence_length`: an integer type, the desired final sequence length, *must be a multiple of 128 given the binned nature of labels<br>
203
 
204
  Input: a genomic nucleotide sequence<br>
205
  Output: a variable length vector depending on the requested sequence length [requested_sequence_length / 128, 50]
 
208
  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
209
  set with the train set and use cross validation to select a new train and validation set from this combined set.
210
 
 
 
 
 
 
211
 
212
  ---
213
 
214
+ ### 5. Bulk RNA Expression
215
+ Gene expression involves the process by which information encoded in a gene directs the synthesis of a functional gene product, typically a protein, through transcription and translation.
216
+ Transcriptional regulation determines the amount of mRNA produced, which is then translated into proteins. Developing a model that can predict RNA expression levels solely from sequence
217
+ data is crucial for advancing our understanding of gene regulation, elucidating disease mechanisms, and identifying functional sequence variants.
218
 
219
  #### Source
220
  Original data comes from GTEx. We use processed data files from the [ExPecto paper](https://www.nature.com/articles/s41588-018-0160-6) found
221
  [here](https://github.com/FunctionLab/ExPecto/tree/master/resources). Sequence data originates from the GRCh37/hg19 genome assembly.
222
 
223
  #### Data Processing
224
+ The authors of ExPecto determined representative TSS for Pol II transcribed genes
225
+ based on quantification of CAGE reads from the FANTOM5 project. The specific procedure they used is as
226
+ follows, a CAGE peak was associated to a GENCODE gene if it was withing 1000 bps from a
227
+ GENCODE v24 annotated TSS. The most abundant CAGE peak for each gene was then selected
228
+ as the representative TSS. When no CAGE peak could be assigned to a gene, the annotated gene
229
+ start position was used as the representative TSS. We log(1 + x) normalized then standardized the
230
+ RNA-seq counts before training models. A list of names of tissues corresponding to
231
+ the labels can be found here: `bulk_rna_expression/label_mapping.csv`. *Note the
232
+ data in this repository for this task has already been log(1+x) normalized and
233
+ standardized to mean 0 and unit variance.
234
 
235
  #### Task Structure
236
 
237
  Type: Multi-variable regression<br>
238
 
239
  Task Args:<br>
240
+ `sequence_length`: an integer type, the desired final sequence length<br>
241
 
242
  Input: a genomic nucleotide sequence centered around the CAGE representative trancription start site<br>
243
  Output: a 218 length vector of continuous values corresponding to the bulk RNA expression levels in 218 different tissue types
 
246
  Train: chromosomes 1-7,9-22,X,Y<br>
247
  Test: chromosome 8
248
 
 
 
 
 
 
249
  ---
250
+ ### 6. Chromatin Features
251
+ Predicting chromatin features, such as histone marks and DNA accessibility, is crucial for understanding gene regulation, as these features indicate chromatin state and are essential for transcription activation.
 
252
 
253
  #### Source
254
+ 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.
255
+ Sequence data originates from the GRCh37/hg19 genome assembly.
256
 
257
  #### Data Processing
258
+ The authors of DeepSea processed the data by chunking the human genome
259
+ into 200 bp bins where for each bin labels were determined for hundreds of different chromatin
260
+ features. Only bins with at least one transcription factor binding event were
261
+ considered for the dataset. If the bin overlapped with a peak region of the specific
262
+ chromatin profile by more than half of the
263
+ sequence, a positive label was assigned. DNA sequences were obtained from the human reference
264
+ genome assembly GRCh37. To make the dataset more accessible, we randomly sub-sampled the
265
+ chromatin profiles from 125 to 20 tracks for the histones dataset and from 104 to 20 tracks for the
266
+ DNA accessibility dataset.
267
 
268
  #### Task Structure
269
 
270
+ Type: Multi-label binary classification
271
 
272
  Task Args:<br>
273
+ `sequence_length`: an integer type, the desired final sequence length<br>
274
+ `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)
275
 
276
+ Input: a genomic nucleotide sequence centered on the 200 base pair bin that is associated with the labels<br>
277
+ Output: a vector of length 20 with binary entries
278
 
279
  #### Splits
280
+ Train set: chromosomes 1-7,10-22<br>
281
+ Test set: chromosomes 8,9
282
+
283
+ ---
284
+ ### 7. Regulatory Elements
285
+ Cis-regulatory elements, such as promoters and enhancers, control the spatial and temporal expression of genes.
286
+ These elements are essential for understanding gene regulation mechanisms and how genetic variations can lead to differences in gene expression.
287
+
288
+ #### Source
289
+ Original data annotations to build labels came from the Search Candidate cis-Regulatory Elements by ENCODE project. Sequence data originates from the GRCh38
290
+ genome assembly.
291
 
292
+ #### Data Processing
293
+ The data is processed as follows, we break the human
294
+ reference genome into 200 bp non-overlapping chunks. If the 200 bp chunk overlaps by at least 50%
295
+ or more with a contiguous region from the set of annotated cis-regulatory elements (promoters or
296
+ enhancers), we label them as positive, else the chunk is labeled as negative. The resulting dataset
297
+ was composed of ∼15M negative samples and ∼50k positive promoter samples and ∼1M positive
298
+ enhancer samples. We randomly sub-sampled the negative set to 1M samples, and kept
299
+ all positive
300
+ samples, to make this dataset more manageable in size.
301
+
302
+ #### Task Structure
303
+
304
+ Type: Binary classification
305
+
306
+ Task Args:<br>
307
+ `sequence_length`: an integer type, the desired final sequence length<br>
308
+ `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)
309
+
310
+ Input: a genomic nucleotide sequence centered on the 200 base pair bin that is associated with the label<br>
311
+ Output: a single binary value
312
+
313
+ #### Splits
314
+ Train set: chromosomes 1-7,10-22<br>
315
+ Test set: chromosomes 8,9
chromatin_features/histones_and_dnase.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6cc04691aca70c876018f15463ba697ddd790af8acda7bbdf14417a3032d153
3
+ size 356382794
chromatin_features/histones_and_dnase_subset.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5a63d4067cd0f1da65aecb6b1ed0a2e1beb19a104f61bfc06ca401b0c14dc14
3
+ size 47732536
genomics-long-range-benchmark.py CHANGED
@@ -14,14 +14,12 @@ import pandas as pd
14
  from datasets import DatasetInfo
15
  from pyfaidx import Fasta
16
  from abc import ABC, abstractmethod
17
- from Bio.Seq import Seq
18
- from Bio import SeqIO
19
- import pysam
20
 
21
  """
22
- --------------------------------------------------------------------------------------------
23
  Reference Genome URLS:
24
- -------------------------------------------------------------------------------------------
25
  """
26
  H38_REFERENCE_GENOME_URL = (
27
  "https://hgdownload.soe.ucsc.edu/goldenPath/hg38/bigZips/" "hg38.fa.gz"
@@ -31,9 +29,9 @@ H19_REFERENCE_GENOME_URL = (
31
  )
32
 
33
  """
34
- --------------------------------------------------------------------------------------------
35
  Task Specific Handlers:
36
- -------------------------------------------------------------------------------------------
37
  """
38
 
39
  class GenomicLRATaskHandler(ABC):
@@ -97,8 +95,8 @@ class GenomicLRATaskHandler(ABC):
97
 
98
  def download_and_extract_gz(self, file_url, cache_dir_root):
99
  """
100
- Downloads and extracts a gz file into the given cache directory. Returns the full file path
101
- of the extracted gz file.
102
  Args:
103
  file_url: url of the gz file to be downloaded and extracted.
104
  cache_dir_root: Directory to extract file into.
@@ -138,29 +136,30 @@ class CagePredictionHandler(GenomicLRATaskHandler):
138
  50,
139
  ) # 50 is a subset of CAGE tracks from the original enformer dataset
140
  NPZ_SPLIT = 1000 # number of files per npz file.
141
- NUM_BP_PER_BIN = 128 # number of base pairs per bin in labels
142
 
143
  def __init__(self, sequence_length=DEFAULT_LENGTH, **kwargs):
144
  """
145
  Creates a new handler for the CAGE task.
146
  Args:
147
- sequence_length: allows for increasing sequence context. Sequence length must be a multiple of 128 to align with binned labels. Note: increasing
148
- sequence length may decrease the number of usable samples.
 
149
  """
150
  self.reference_genome = None
151
  self.coordinate_csv_file = None
152
  self.target_files_by_split = {}
153
 
 
154
  assert (sequence_length // 128) % 2 == 0, (
155
- f"Requested sequence length must be an even multuple of 128 to align with the binned labels."
 
156
  )
157
 
158
  self.sequence_length = sequence_length
159
 
160
  if self.sequence_length < self.DEFAULT_LENGTH:
161
-
162
- self.TARGET_SHAPE = (self.sequence_length//128,50)
163
-
164
 
165
  def get_info(self, description: str) -> DatasetInfo:
166
  """
@@ -174,7 +173,11 @@ class CagePredictionHandler(GenomicLRATaskHandler):
174
  # array of sequence length x num_labels
175
  "labels": datasets.Array2D(shape=self.TARGET_SHAPE, dtype="float32"),
176
  # chromosome number
177
- "chromosome":datasets.Value(dtype="string")
 
 
 
 
178
  }
179
  )
180
  return datasets.DatasetInfo(
@@ -192,7 +195,7 @@ class CagePredictionHandler(GenomicLRATaskHandler):
192
  """
193
 
194
  # Manually download the reference genome since there are difficulties when
195
- # streaming the dataset
196
  reference_genome_file = self.download_and_extract_gz(
197
  H38_REFERENCE_GENOME_URL, cache_dir_root
198
  )
@@ -225,7 +228,6 @@ class CagePredictionHandler(GenomicLRATaskHandler):
225
  self.target_files_by_split["test"] = test_file_dict
226
  self.target_files_by_split["validation"] = valid_file_dict
227
 
228
-
229
  return [
230
  datasets.SplitGenerator(
231
  name=datasets.Split.TRAIN,
@@ -241,7 +243,6 @@ class CagePredictionHandler(GenomicLRATaskHandler):
241
  ),
242
  ]
243
 
244
-
245
  def generate_examples(self, split):
246
  """
247
  A generator which produces examples for the given split, each with a sequence
@@ -250,24 +251,28 @@ class CagePredictionHandler(GenomicLRATaskHandler):
250
  """
251
 
252
  target_files = self.target_files_by_split[split]
253
- sequence_length = self.sequence_length
254
 
255
  key = 0
256
  coordinates_dataframe = pd.read_csv(self.coordinate_csv_file)
257
  filtered = coordinates_dataframe[coordinates_dataframe["split"] == split]
258
  for sequential_idx, row in filtered.iterrows():
259
  start, stop = int(row["start"]) - 1, int(
260
- row["stop"]) - 1 # -1 since vcf coords are 1-based
261
 
262
  chromosome = row['chrom']
263
-
264
- padded_sequence = pad_sequence(
265
  chromosome=self.reference_genome[chromosome],
266
  start=start,
267
- sequence_length=sequence_length,
268
  end=stop,
 
269
  )
270
 
 
 
 
 
271
  # floor npy_idx to the nearest 1000
272
  npz_file = np.load(
273
  target_files[int((row["npy_idx"] // self.NPZ_SPLIT) * self.NPZ_SPLIT)]
@@ -277,21 +282,22 @@ class CagePredictionHandler(GenomicLRATaskHandler):
277
  split == "validation"
278
  ): # npy files are keyed by ["train", "test", "valid"]
279
  split = "valid"
280
- targets = npz_file[f"target-{split}-{row['npy_idx']}.npy"][0] # select 0 since there is extra dimension
281
-
282
-
283
  # subset the targets if sequence length is smaller than 114688 (
284
  # DEFAULT_LENGTH)
285
  if self.sequence_length < self.DEFAULT_LENGTH:
286
  idx_diff = (self.DEFAULT_LENGTH - self.sequence_length) // 2 // 128
287
  targets = targets[idx_diff:-idx_diff]
288
 
289
-
290
  if padded_sequence:
291
  yield key, {
292
  "labels": targets,
293
  "sequence": standardize_sequence(padded_sequence),
294
- "chromosome": re.sub("chr","",chromosome)
 
 
295
  }
296
  key += 1
297
 
@@ -325,7 +331,7 @@ class BulkRnaExpressionHandler(GenomicLRATaskHandler):
325
  Handler for the Bulk RNA Expression task.
326
  """
327
 
328
- DEFAULT_LENGTH = 114688
329
 
330
  def __init__(self, sequence_length=DEFAULT_LENGTH, **kwargs):
331
  """
@@ -351,7 +357,9 @@ class BulkRnaExpressionHandler(GenomicLRATaskHandler):
351
  # list of expression values in each tissue
352
  "labels": datasets.Sequence(datasets.Value("float32")),
353
  # chromosome number
354
- "chromosome":datasets.Value(dtype="string")
 
 
355
  }
356
  )
357
  return datasets.DatasetInfo(
@@ -368,7 +376,7 @@ class BulkRnaExpressionHandler(GenomicLRATaskHandler):
368
  The Bulk RNA Expression dataset requires the reference hg19 genome, coordinate
369
  csv file,and label csv file to be saved.
370
  """
371
- # Manually download the reference genome since there are difficulties when streaming
372
  reference_genome_file = self.download_and_extract_gz(
373
  H19_REFERENCE_GENOME_URL, cache_dir_root
374
  )
@@ -398,7 +406,7 @@ class BulkRnaExpressionHandler(GenomicLRATaskHandler):
398
  key = 0
399
  for idx, coordinates_row in coordinates_split_df.iterrows():
400
  start = coordinates_row[
401
- "CAGE_representative_TSS"] - 1 # -1 since vcf coords are 1-based
402
 
403
  chromosome = coordinates_row["chrom"]
404
  labels_row = labels_df.loc[idx].values
@@ -412,21 +420,22 @@ class BulkRnaExpressionHandler(GenomicLRATaskHandler):
412
  yield key, {
413
  "labels": labels_row,
414
  "sequence": standardize_sequence(padded_sequence),
415
- "chromosome":re.sub("chr","",chromosome)
 
416
  }
417
  key += 1
418
 
419
 
420
- class VariantEffectPredictionHandler(GenomicLRATaskHandler):
421
  """
422
- Handler for the Variant Effect Prediction task.
423
  """
424
 
425
- DEFAULT_LENGTH = 114688
426
 
427
  def __init__(self, sequence_length=DEFAULT_LENGTH, **kwargs):
428
  """
429
- Creates a new handler for the Variant Effect Prediction Task.
430
  Args:
431
  sequence_length: Length of the sequence to pad around the SNP position
432
 
@@ -436,9 +445,9 @@ class VariantEffectPredictionHandler(GenomicLRATaskHandler):
436
 
437
  def get_info(self, description: str) -> DatasetInfo:
438
  """
439
- Returns the DatasetInfo for the Variant Effect Prediction dataset. Each example
440
- includes a genomic sequence with the reference allele as well as the genomic sequence with the alternative allele,
441
- and a binary label.
442
  """
443
  features = datasets.Features(
444
  {
@@ -451,8 +460,10 @@ class VariantEffectPredictionHandler(GenomicLRATaskHandler):
451
  "tissue": datasets.Value(dtype="string"),
452
  # chromosome number
453
  "chromosome": datasets.Value(dtype="string"),
 
 
454
  # distance to nearest tss
455
- "distance_to_nearest_tss":datasets.Value(dtype="int32")
456
  }
457
  )
458
 
@@ -478,7 +489,7 @@ class VariantEffectPredictionHandler(GenomicLRATaskHandler):
478
 
479
  self.reference_genome = Fasta(reference_genome_file, one_based_attributes=False)
480
  self.coordinates_labels_csv_file = dl_manager.download_and_extract(
481
- f"variant_effect_gene_expression/All_Tissues.csv"
482
  )
483
 
484
  return super().split_generators(dl_manager, cache_dir_root)
@@ -496,7 +507,7 @@ class VariantEffectPredictionHandler(GenomicLRATaskHandler):
496
 
497
  key = 0
498
  for idx, row in coordinates_split_df.iterrows():
499
- start = row["POS"] - 1 # sub 1 to create idx since vcf coords are 1-based
500
  alt_allele = row["ALT"]
501
  label = row["label"]
502
  tissue = row['tissue']
@@ -513,8 +524,8 @@ class VariantEffectPredictionHandler(GenomicLRATaskHandler):
513
 
514
  # only if a valid sequence returned
515
  if ref_forward:
516
- # Mutate sequence with the alt allele at the SNP position, which is always
517
- # centered in the string returned from pad_sequence
518
  alt_forward = list(ref_forward)
519
  alt_forward[self.sequence_length // 2] = alt_allele
520
  alt_forward = "".join(alt_forward)
@@ -525,14 +536,354 @@ class VariantEffectPredictionHandler(GenomicLRATaskHandler):
525
  "chromosome": re.sub("chr", "", chromosome),
526
  "ref_forward_sequence": standardize_sequence(ref_forward),
527
  "alt_forward_sequence": standardize_sequence(alt_forward),
528
- "distance_to_nearest_tss": distance
 
529
  }
530
  key += 1
531
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
532
  """
533
- --------------------------------------------------------------------------------------------
534
  Dataset loader:
535
- -------------------------------------------------------------------------------------------
536
  """
537
 
538
  _DESCRIPTION = """
@@ -542,7 +893,13 @@ Dataset for benchmark of genomic deep learning models.
542
  _TASK_HANDLERS = {
543
  "cage_prediction": CagePredictionHandler,
544
  "bulk_rna_expression": BulkRnaExpressionHandler,
545
- "variant_effect_gene_expression": VariantEffectPredictionHandler,
 
 
 
 
 
 
546
  }
547
 
548
 
@@ -558,7 +915,7 @@ class GenomicsLRAConfig(datasets.BuilderConfig):
558
  **kwargs: keyword arguments forwarded to super.
559
  """
560
  super().__init__()
561
- self.handler = _TASK_HANDLERS[task_name](task_name=task_name,**kwargs)
562
 
563
 
564
  # DatasetBuilder
@@ -592,9 +949,9 @@ class GenomicsLRATasks(datasets.GeneratorBasedBuilder):
592
 
593
 
594
  """
595
- --------------------------------------------------------------------------------------------
596
  Global Utils:
597
- -------------------------------------------------------------------------------------------
598
  """
599
 
600
 
@@ -613,7 +970,8 @@ def standardize_sequence(sequence: str):
613
  return sequence
614
 
615
 
616
- def pad_sequence(chromosome, start, sequence_length, end=None, negative_strand=False):
 
617
  """
618
  Extends a given sequence to length sequence_length. If
619
  padding to the given length is outside the gene, returns
@@ -625,7 +983,8 @@ def pad_sequence(chromosome, start, sequence_length, end=None, negative_strand=F
625
  remainder is added to the end of the sequence.
626
  end: End index of original sequence. If no end is specified, it creates a
627
  centered sequence around the start index.
628
- negative_strand: If negative_strand, returns the reverse compliment of the sequence
 
629
  """
630
  if end:
631
  pad = (sequence_length - (end - start)) // 2
@@ -639,5 +998,12 @@ def pad_sequence(chromosome, start, sequence_length, end=None, negative_strand=F
639
  if start < 0 or end >= len(chromosome):
640
  return
641
  if negative_strand:
 
 
 
642
  return chromosome[start:end].reverse.complement.seq
 
 
 
 
643
  return chromosome[start:end].seq
 
14
  from datasets import DatasetInfo
15
  from pyfaidx import Fasta
16
  from abc import ABC, abstractmethod
17
+
 
 
18
 
19
  """
20
+ ----------------------------------------------------------------------------------------
21
  Reference Genome URLS:
22
+ ----------------------------------------------------------------------------------------
23
  """
24
  H38_REFERENCE_GENOME_URL = (
25
  "https://hgdownload.soe.ucsc.edu/goldenPath/hg38/bigZips/" "hg38.fa.gz"
 
29
  )
30
 
31
  """
32
+ ----------------------------------------------------------------------------------------
33
  Task Specific Handlers:
34
+ ----------------------------------------------------------------------------------------
35
  """
36
 
37
  class GenomicLRATaskHandler(ABC):
 
95
 
96
  def download_and_extract_gz(self, file_url, cache_dir_root):
97
  """
98
+ Downloads and extracts a gz file into the given cache directory. Returns the
99
+ full file path of the extracted gz file.
100
  Args:
101
  file_url: url of the gz file to be downloaded and extracted.
102
  cache_dir_root: Directory to extract file into.
 
136
  50,
137
  ) # 50 is a subset of CAGE tracks from the original enformer dataset
138
  NPZ_SPLIT = 1000 # number of files per npz file.
139
+ NUM_BP_PER_BIN = 128 # number of base pairs per bin in labels
140
 
141
  def __init__(self, sequence_length=DEFAULT_LENGTH, **kwargs):
142
  """
143
  Creates a new handler for the CAGE task.
144
  Args:
145
+ sequence_length: allows for increasing sequence context. Sequence length
146
+ must be an even multiple of 128 to align with binned labels. Note:
147
+ increasing sequence length may decrease the number of usable samples.
148
  """
149
  self.reference_genome = None
150
  self.coordinate_csv_file = None
151
  self.target_files_by_split = {}
152
 
153
+
154
  assert (sequence_length // 128) % 2 == 0, (
155
+ f"Requested sequence length must be an even multuple of 128 to align "
156
+ f"with the binned labels."
157
  )
158
 
159
  self.sequence_length = sequence_length
160
 
161
  if self.sequence_length < self.DEFAULT_LENGTH:
162
+ self.TARGET_SHAPE = (self.sequence_length // 128, 50)
 
 
163
 
164
  def get_info(self, description: str) -> DatasetInfo:
165
  """
 
173
  # array of sequence length x num_labels
174
  "labels": datasets.Array2D(shape=self.TARGET_SHAPE, dtype="float32"),
175
  # chromosome number
176
+ "chromosome": datasets.Value(dtype="string"),
177
+ # start
178
+ "labels_start": datasets.Value(dtype="int32"),
179
+ # stop
180
+ "labels_stop": datasets.Value(dtype="int32")
181
  }
182
  )
183
  return datasets.DatasetInfo(
 
195
  """
196
 
197
  # Manually download the reference genome since there are difficulties when
198
+ # streaming
199
  reference_genome_file = self.download_and_extract_gz(
200
  H38_REFERENCE_GENOME_URL, cache_dir_root
201
  )
 
228
  self.target_files_by_split["test"] = test_file_dict
229
  self.target_files_by_split["validation"] = valid_file_dict
230
 
 
231
  return [
232
  datasets.SplitGenerator(
233
  name=datasets.Split.TRAIN,
 
243
  ),
244
  ]
245
 
 
246
  def generate_examples(self, split):
247
  """
248
  A generator which produces examples for the given split, each with a sequence
 
251
  """
252
 
253
  target_files = self.target_files_by_split[split]
 
254
 
255
  key = 0
256
  coordinates_dataframe = pd.read_csv(self.coordinate_csv_file)
257
  filtered = coordinates_dataframe[coordinates_dataframe["split"] == split]
258
  for sequential_idx, row in filtered.iterrows():
259
  start, stop = int(row["start"]) - 1, int(
260
+ row["stop"]) - 1 # -1 since coords are 1-based
261
 
262
  chromosome = row['chrom']
263
+
264
+ padded_sequence,new_start,new_stop = pad_sequence(
265
  chromosome=self.reference_genome[chromosome],
266
  start=start,
267
+ sequence_length=self.sequence_length,
268
  end=stop,
269
+ return_new_start_stop=True
270
  )
271
 
272
+ if self.sequence_length >= self.DEFAULT_LENGTH:
273
+ new_start = start
274
+ new_stop = stop
275
+
276
  # floor npy_idx to the nearest 1000
277
  npz_file = np.load(
278
  target_files[int((row["npy_idx"] // self.NPZ_SPLIT) * self.NPZ_SPLIT)]
 
282
  split == "validation"
283
  ): # npy files are keyed by ["train", "test", "valid"]
284
  split = "valid"
285
+ targets = npz_file[f"target-{split}-{row['npy_idx']}.npy"][
286
+ 0] # select 0 since there is extra dimension
287
+
288
  # subset the targets if sequence length is smaller than 114688 (
289
  # DEFAULT_LENGTH)
290
  if self.sequence_length < self.DEFAULT_LENGTH:
291
  idx_diff = (self.DEFAULT_LENGTH - self.sequence_length) // 2 // 128
292
  targets = targets[idx_diff:-idx_diff]
293
 
 
294
  if padded_sequence:
295
  yield key, {
296
  "labels": targets,
297
  "sequence": standardize_sequence(padded_sequence),
298
+ "chromosome": re.sub("chr", "", chromosome),
299
+ "labels_start": new_start,
300
+ "labels_stop": new_stop
301
  }
302
  key += 1
303
 
 
331
  Handler for the Bulk RNA Expression task.
332
  """
333
 
334
+ DEFAULT_LENGTH = 100000
335
 
336
  def __init__(self, sequence_length=DEFAULT_LENGTH, **kwargs):
337
  """
 
357
  # list of expression values in each tissue
358
  "labels": datasets.Sequence(datasets.Value("float32")),
359
  # chromosome number
360
+ "chromosome": datasets.Value(dtype="string"),
361
+ # position
362
+ "position": datasets.Value(dtype="int32"),
363
  }
364
  )
365
  return datasets.DatasetInfo(
 
376
  The Bulk RNA Expression dataset requires the reference hg19 genome, coordinate
377
  csv file,and label csv file to be saved.
378
  """
379
+
380
  reference_genome_file = self.download_and_extract_gz(
381
  H19_REFERENCE_GENOME_URL, cache_dir_root
382
  )
 
406
  key = 0
407
  for idx, coordinates_row in coordinates_split_df.iterrows():
408
  start = coordinates_row[
409
+ "CAGE_representative_TSS"] - 1 # -1 since coords are 1-based
410
 
411
  chromosome = coordinates_row["chrom"]
412
  labels_row = labels_df.loc[idx].values
 
420
  yield key, {
421
  "labels": labels_row,
422
  "sequence": standardize_sequence(padded_sequence),
423
+ "chromosome": re.sub("chr", "", chromosome),
424
+ "position": coordinates_row["CAGE_representative_TSS"]
425
  }
426
  key += 1
427
 
428
 
429
+ class VariantEffectCausalEqtl(GenomicLRATaskHandler):
430
  """
431
+ Handler for the Variant Effect Causal eQTL task.
432
  """
433
 
434
+ DEFAULT_LENGTH = 100000
435
 
436
  def __init__(self, sequence_length=DEFAULT_LENGTH, **kwargs):
437
  """
438
+ Creates a new handler for the Variant Effect Causal eQTL Task.
439
  Args:
440
  sequence_length: Length of the sequence to pad around the SNP position
441
 
 
445
 
446
  def get_info(self, description: str) -> DatasetInfo:
447
  """
448
+ Returns the DatasetInfo for the Variant Effect Causal eQTL dataset. Each example
449
+ includes a genomic sequence with the reference allele as well as the genomic
450
+ sequence with the alternative allele, and a binary label.
451
  """
452
  features = datasets.Features(
453
  {
 
460
  "tissue": datasets.Value(dtype="string"),
461
  # chromosome number
462
  "chromosome": datasets.Value(dtype="string"),
463
+ # variant position
464
+ "position": datasets.Value(dtype="int32"),
465
  # distance to nearest tss
466
+ "distance_to_nearest_tss": datasets.Value(dtype="int32")
467
  }
468
  )
469
 
 
489
 
490
  self.reference_genome = Fasta(reference_genome_file, one_based_attributes=False)
491
  self.coordinates_labels_csv_file = dl_manager.download_and_extract(
492
+ f"variant_effect_causal_eqtl/All_Tissues.csv"
493
  )
494
 
495
  return super().split_generators(dl_manager, cache_dir_root)
 
507
 
508
  key = 0
509
  for idx, row in coordinates_split_df.iterrows():
510
+ start = row["POS"] - 1 # sub 1 to create idx since coords are 1-based
511
  alt_allele = row["ALT"]
512
  label = row["label"]
513
  tissue = row['tissue']
 
524
 
525
  # only if a valid sequence returned
526
  if ref_forward:
527
+ # Mutate sequence with the alt allele at the SNP position,
528
+ # which is always centered in the string returned from pad_sequence
529
  alt_forward = list(ref_forward)
530
  alt_forward[self.sequence_length // 2] = alt_allele
531
  alt_forward = "".join(alt_forward)
 
536
  "chromosome": re.sub("chr", "", chromosome),
537
  "ref_forward_sequence": standardize_sequence(ref_forward),
538
  "alt_forward_sequence": standardize_sequence(alt_forward),
539
+ "distance_to_nearest_tss": distance,
540
+ "position": row["POS"]
541
  }
542
  key += 1
543
 
544
+
545
+ class VariantEffectPathogenicHandler(GenomicLRATaskHandler):
546
+ """
547
+ Handler for the Variant Effect Pathogenic Prediction tasks.
548
+ """
549
+
550
+ DEFAULT_LENGTH = 100000
551
+
552
+ def __init__(self, sequence_length=DEFAULT_LENGTH, task_name=None, subset=False,
553
+ **kwargs):
554
+ """
555
+ Creates a new handler for the Variant Effect Pathogenic Tasks.
556
+ Args:
557
+ sequence_length: Length of the sequence to pad around the SNP position
558
+ subset: Whether to return a pre-determined subset of the data.
559
+
560
+ """
561
+ self.sequence_length = sequence_length
562
+
563
+ if task_name == 'variant_effect_pathogenic_clinvar':
564
+ self.data_file_name = "variant_effect_pathogenic/vep_pathogenic_coding.csv"
565
+ elif task_name == 'variant_effect_pathogenic_omim':
566
+ self.data_file_name = "variant_effect_pathogenic/" \
567
+ "vep_pathogenic_non_coding_subset.csv" \
568
+ if subset else "variant_effect_pathogenic/vep_pathogenic_non_coding.csv"
569
+
570
+ def get_info(self, description: str) -> DatasetInfo:
571
+ """
572
+ Returns the DatasetInfo for the Variant Effect Pathogenic datasets. Each example
573
+ includes a genomic sequence with the reference allele as well as the genomic
574
+ sequence with the alternative allele, and a binary label.
575
+ """
576
+ features = datasets.Features(
577
+ {
578
+ # DNA sequence
579
+ "ref_forward_sequence": datasets.Value("string"),
580
+ "alt_forward_sequence": datasets.Value("string"),
581
+ # binary label
582
+ "label": datasets.Value(dtype="int8"),
583
+ # chromosome number
584
+ "chromosome": datasets.Value(dtype="string"),
585
+ # position
586
+ "position": datasets.Value(dtype="int32")
587
+ }
588
+ )
589
+
590
+ return datasets.DatasetInfo(
591
+ # This is the description that will appear on the datasets page.
592
+ description=description,
593
+ # This defines the different columns of the dataset and their types
594
+ features=features,
595
+ )
596
+
597
+ def split_generators(self, dl_manager, cache_dir_root):
598
+ """
599
+ Separates files by split and stores filenames in instance variables.
600
+ The variant effect prediction datasets require the reference hg38 genome and
601
+ coordinates_labels_csv_file to be saved.
602
+ """
603
+
604
+ reference_genome_file = self.download_and_extract_gz(
605
+ H38_REFERENCE_GENOME_URL, cache_dir_root
606
+ )
607
+
608
+ self.reference_genome = Fasta(reference_genome_file, one_based_attributes=False)
609
+ self.coordinates_labels_csv_file = dl_manager.download_and_extract(
610
+ self.data_file_name)
611
+
612
+ if 'non_coding' in self.data_file_name:
613
+ return [
614
+ datasets.SplitGenerator(
615
+ name=datasets.Split.TEST,
616
+ gen_kwargs={"handler": self, "split": "test"}
617
+ ), ]
618
+ else:
619
+ return super().split_generators(dl_manager, cache_dir_root)
620
+
621
+ def generate_examples(self, split):
622
+ """
623
+ A generator which produces examples each with ref/alt allele
624
+ and corresponding binary label. The sequences are extended to
625
+ the desired sequence length and standardized before returning.
626
+ """
627
+
628
+ coordinates_df = pd.read_csv(self.coordinates_labels_csv_file)
629
+ coordinates_split_df = coordinates_df[coordinates_df["split"] == split]
630
+
631
+ key = 0
632
+ for idx, row in coordinates_split_df.iterrows():
633
+ start = row["POS"] - 1 # sub 1 to create idx since coords are 1-based
634
+ alt_allele = row["ALT"]
635
+ label = row["INT_LABEL"]
636
+ chromosome = row["CHROM"]
637
+
638
+ # get reference forward sequence
639
+ ref_forward = pad_sequence(
640
+ chromosome=self.reference_genome[chromosome],
641
+ start=start,
642
+ sequence_length=self.sequence_length,
643
+ negative_strand=False,
644
+ )
645
+
646
+ # only if a valid sequence returned
647
+ if ref_forward:
648
+ # Mutate sequence with the alt allele at the SNP position,
649
+ # which is always centered in the string returned from pad_sequence
650
+ alt_forward = list(ref_forward)
651
+ alt_forward[self.sequence_length // 2] = alt_allele
652
+ alt_forward = "".join(alt_forward)
653
+
654
+ yield key, {
655
+ "label": label,
656
+ "chromosome": re.sub("chr", "", chromosome),
657
+ "ref_forward_sequence": standardize_sequence(ref_forward),
658
+ "alt_forward_sequence": standardize_sequence(alt_forward),
659
+ "position": row['POS']
660
+ }
661
+ key += 1
662
+
663
+
664
+ class ChromatinFeaturesHandler(GenomicLRATaskHandler):
665
+ """
666
+ Handler for the histone marks and DNA accessibility tasks also referred to
667
+ collectively as Chromatin features.
668
+ """
669
+
670
+ DEFAULT_LENGTH = 100000
671
+
672
+ def __init__(self, task_name=None, sequence_length=DEFAULT_LENGTH, subset=False,
673
+ **kwargs):
674
+ """
675
+ Creates a new handler for the Deep Sea Histone and DNase tasks.
676
+ Args:
677
+ sequence_length: Length of the sequence around and including the
678
+ annotated 200bp bin
679
+ subset: Whether to return a pre-determined subset of the entire dataset.
680
+
681
+ """
682
+ self.sequence_length = sequence_length
683
+
684
+ if sequence_length < 200:
685
+ raise ValueError(
686
+ 'Sequence length for this task must be greater or equal to 200 bp')
687
+
688
+ if 'histone' in task_name:
689
+ self.label_name = 'HISTONES'
690
+ elif 'dna' in task_name:
691
+ self.label_name = 'DNASE'
692
+
693
+ self.data_file_name = "chromatin_features/histones_and_dnase_subset.csv" if \
694
+ subset else "chromatin_features/histones_and_dnase.csv"
695
+
696
+ def get_info(self, description: str) -> DatasetInfo:
697
+ """
698
+ Returns the DatasetInfo for the histone marks and dna accessibility datasets.
699
+ Each example includes a genomic sequence and a list of label values.
700
+ """
701
+ features = datasets.Features(
702
+ {
703
+ # DNA sequence
704
+ "sequence": datasets.Value("string"),
705
+ # list of binary chromatin marks
706
+ "labels": datasets.Sequence(datasets.Value("int8")),
707
+ # chromosome number
708
+ "chromosome": datasets.Value(dtype="string"),
709
+ # starting position in genome which corresponds to label
710
+ "label_start": datasets.Value(dtype="int32"),
711
+ # end position in genome which corresponds to label
712
+ "label_stop": datasets.Value(dtype="int32"),
713
+ }
714
+ )
715
+ return datasets.DatasetInfo(
716
+ # This is the description that will appear on the datasets page.
717
+ description=description,
718
+ # This defines the different columns of the dataset and their types
719
+ features=features,
720
+
721
+ )
722
+
723
+ def split_generators(self, dl_manager, cache_dir_root):
724
+ """
725
+ Separates files by split and stores filenames in instance variables.
726
+ The histone marks and dna accessibility datasets require the reference hg19
727
+ genome and coordinate csv file to be saved.
728
+ """
729
+ reference_genome_file = self.download_and_extract_gz(
730
+ H19_REFERENCE_GENOME_URL, cache_dir_root
731
+ )
732
+ self.reference_genome = Fasta(reference_genome_file, one_based_attributes=False)
733
+
734
+ self.coordinate_csv_file = dl_manager.download_and_extract(self.data_file_name)
735
+
736
+ return super().split_generators(dl_manager, cache_dir_root)
737
+
738
+ def generate_examples(self, split):
739
+ """
740
+ A generator which produces examples for the given split, each with a sequence
741
+ and the corresponding labels. The sequences are padded to the correct sequence
742
+ length and standardized before returning.
743
+ """
744
+ coordinates_df = pd.read_csv(self.coordinate_csv_file)
745
+ coordinates_split_df = coordinates_df[coordinates_df["split"] == split]
746
+
747
+ key = 0
748
+ for idx, coordinates_row in coordinates_split_df.iterrows():
749
+ start = coordinates_row['POS'] - 1 # -1 since saved coords are 1-based
750
+ chromosome = coordinates_row["CHROM"]
751
+
752
+ # literal eval used since lists are saved as strings in csv
753
+ labels_row = literal_eval(coordinates_row[self.label_name])
754
+
755
+ padded_sequence = pad_sequence(
756
+ chromosome=self.reference_genome[chromosome],
757
+ start=start,
758
+ sequence_length=self.sequence_length,
759
+ )
760
+ if padded_sequence:
761
+ yield key, {
762
+ "labels": labels_row,
763
+ "sequence": standardize_sequence(padded_sequence),
764
+ "chromosome": re.sub("chr", "", chromosome),
765
+ "label_start": coordinates_row['POS']-100,
766
+ "label_stop": coordinates_row['POS'] + 99,
767
+ }
768
+ key += 1
769
+
770
+
771
+ class RegulatoryElementHandler(GenomicLRATaskHandler):
772
+ """
773
+ Handler for the Regulatory Element Prediction tasks.
774
+ """
775
+ DEFAULT_LENGTH = 100000
776
+
777
+ def __init__(self, task_name=None, sequence_length=DEFAULT_LENGTH, subset=False,
778
+ **kwargs):
779
+ """
780
+ Creates a new handler for the Regulatory Element Prediction tasks.
781
+ Args:
782
+ sequence_length: Length of the sequence around the element/non-element
783
+ subset: Whether to return a pre-determined subset of the entire dataset.
784
+
785
+ """
786
+
787
+ if sequence_length < 200:
788
+ raise ValueError(
789
+ 'Sequence length for this task must be greater or equal to 200 bp')
790
+
791
+ self.sequence_length = sequence_length
792
+
793
+ if 'promoter' in task_name:
794
+ self.data_file_name = 'regulatory_elements/promoter_dataset'
795
+
796
+ elif 'enhancer' in task_name:
797
+ self.data_file_name = 'regulatory_elements/enhancer_dataset'
798
+
799
+ if subset:
800
+ self.data_file_name += '_subset.csv'
801
+ else:
802
+ self.data_file_name += '.csv'
803
+
804
+ def get_info(self, description: str) -> DatasetInfo:
805
+ """
806
+ Returns the DatasetInfo for the Regulatory Element Prediction Tasks.
807
+ Each example includes a genomic sequence and a label.
808
+ """
809
+ features = datasets.Features(
810
+ {
811
+ # DNA sequence
812
+ "sequence": datasets.Value("string"),
813
+ # label corresponding to whether the sequence has
814
+ # the regulatory element of interest or not
815
+ "labels": datasets.Value("int8"),
816
+ # chromosome number
817
+ "chromosome": datasets.Value(dtype="string"),
818
+ # start
819
+ "label_start": datasets.Value(dtype="int32"),
820
+ # stop
821
+ "label_stop": datasets.Value(dtype="int32"),
822
+ }
823
+ )
824
+ return datasets.DatasetInfo(
825
+ # This is the description that will appear on the datasets page.
826
+ description=description,
827
+ # This defines the different columns of the dataset and their types
828
+ features=features,
829
+
830
+ )
831
+
832
+ def split_generators(self, dl_manager, cache_dir_root):
833
+ """
834
+ Separates files by split and stores filenames in instance variables.
835
+ """
836
+ reference_genome_file = self.download_and_extract_gz(
837
+ H38_REFERENCE_GENOME_URL, cache_dir_root
838
+ )
839
+ self.reference_genome = Fasta(reference_genome_file, one_based_attributes=False)
840
+
841
+ self.coordinate_csv_file = dl_manager.download_and_extract(
842
+ self.data_file_name
843
+ )
844
+
845
+ return super().split_generators(dl_manager, cache_dir_root)
846
+
847
+ def generate_examples(self, split):
848
+ """
849
+ A generator which produces examples for the given split, each with a sequence
850
+ and the corresponding label. The sequences are padded to the correct sequence
851
+ length and standardized before returning.
852
+ """
853
+ coordinates_df = pd.read_csv(self.coordinate_csv_file)
854
+
855
+ coordinates_split_df = coordinates_df[coordinates_df["split"] == split]
856
+
857
+ key = 0
858
+ for _, coordinates_row in coordinates_split_df.iterrows():
859
+ start = coordinates_row["START"] - 1 # -1 since vcf coords are 1-based
860
+ end = coordinates_row["STOP"] - 1 # -1 since vcf coords are 1-based
861
+ chromosome = coordinates_row["CHROM"]
862
+
863
+ label = coordinates_row['label']
864
+
865
+ padded_sequence = pad_sequence(
866
+ chromosome=self.reference_genome[chromosome],
867
+ start=start,
868
+ end=end,
869
+ sequence_length=self.sequence_length,
870
+ )
871
+
872
+ if padded_sequence:
873
+ yield key, {
874
+ "labels": label,
875
+ "sequence": standardize_sequence(padded_sequence),
876
+ "chromosome": re.sub("chr", "", chromosome),
877
+ "label_start": coordinates_row["START"],
878
+ "label_stop": coordinates_row["STOP"]
879
+ }
880
+ key += 1
881
+
882
+
883
  """
884
+ ----------------------------------------------------------------------------------------
885
  Dataset loader:
886
+ ----------------------------------------------------------------------------------------
887
  """
888
 
889
  _DESCRIPTION = """
 
893
  _TASK_HANDLERS = {
894
  "cage_prediction": CagePredictionHandler,
895
  "bulk_rna_expression": BulkRnaExpressionHandler,
896
+ "variant_effect_causal_eqtl": VariantEffectCausalEqtl,
897
+ "variant_effect_pathogenic_clinvar": VariantEffectPathogenicHandler,
898
+ "variant_effect_pathogenic_omim": VariantEffectPathogenicHandler,
899
+ "chromatin_features_histone_marks": ChromatinFeaturesHandler,
900
+ "chromatin_features_dna_accessibility": ChromatinFeaturesHandler,
901
+ "regulatory_element_promoter": RegulatoryElementHandler,
902
+ "regulatory_element_enhancer": RegulatoryElementHandler,
903
  }
904
 
905
 
 
915
  **kwargs: keyword arguments forwarded to super.
916
  """
917
  super().__init__()
918
+ self.handler = _TASK_HANDLERS[task_name](task_name=task_name, **kwargs)
919
 
920
 
921
  # DatasetBuilder
 
949
 
950
 
951
  """
952
+ ----------------------------------------------------------------------------------------
953
  Global Utils:
954
+ ----------------------------------------------------------------------------------------
955
  """
956
 
957
 
 
970
  return sequence
971
 
972
 
973
+ def pad_sequence(chromosome, start, sequence_length, end=None, negative_strand=False,
974
+ return_new_start_stop=False):
975
  """
976
  Extends a given sequence to length sequence_length. If
977
  padding to the given length is outside the gene, returns
 
983
  remainder is added to the end of the sequence.
984
  end: End index of original sequence. If no end is specified, it creates a
985
  centered sequence around the start index.
986
+ negative_strand: If negative_strand, returns the reverse compliment of the
987
+ sequence
988
  """
989
  if end:
990
  pad = (sequence_length - (end - start)) // 2
 
998
  if start < 0 or end >= len(chromosome):
999
  return
1000
  if negative_strand:
1001
+ if return_new_start_stop:
1002
+ return chromosome[start:end].reverse.complement.seq ,start, end
1003
+
1004
  return chromosome[start:end].reverse.complement.seq
1005
+
1006
+ if return_new_start_stop:
1007
+ return chromosome[start:end].seq , start, end
1008
+
1009
  return chromosome[start:end].seq
regulatory_elements/enhancer_dataset.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a511212c76aba8c25ecd372531fe78767cbd4065ce72d4899dcb4ba1429750c
3
+ size 66953920
regulatory_elements/enhancer_dataset_subset.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7af8aef4a6c15a4be5d752f8a638f2ab88be42476f9da59732f892273d58078f
3
+ size 11241394
regulatory_elements/promoter_dataset.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ddaaa24bee906885033fb4f75097797c3aed3557b3fb09a202f50d4e99e3a2b
3
+ size 33319573
regulatory_elements/promoter_dataset_subset.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a5ccb2bf002dc504f6c40550acd5a99d2edfa587403d7cf23a1b965df74c51b
3
+ size 7479637
variant_effect_causal_eqtl/All_Tissues.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d85df2b982e57866fb6efcd78dd0316c4005a0fea72322c3fe6c3188772558ae
3
+ size 7081158
variant_effect_gene_expression/All_Tissues.csv DELETED
The diff for this file is too large to render. See raw diff
 
variant_effect_pathogenic/vep_pathogenic_coding.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e75b0b9c374e2cea972cb2c6246c7cc21c1c92e6b97cc597baf20b8d9462acb9
3
+ size 6576847
variant_effect_pathogenic/vep_pathogenic_non_coding.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ab457120792566dcfaeeaf06c9af32c19ddcff4d2f62fb034b95450622f9966
3
+ size 369157498
variant_effect_pathogenic/vep_pathogenic_non_coding_subset.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24127410937b99d41d0ad0b3b1e2e260b27ee20d00695caf6cbb48842a411a1f
3
+ size 33142664