Datasets:

Marks-lab
/

LOL-EVE-UltraRare

language:
  - en
license: mit
task_categories:
  - text-generation
size_categories:
  - 100K<n<1M
configs:
  - config_name: default
    data_files: '*.parquet'
    default: true

LOL-EVE Ultra Rare Variants Dataset

Dataset Description

This dataset contains ultra rare variants from gnomAD for evaluating variant effect prediction models. The dataset includes variants with minor allele frequency (MAF) < 0.001 in promoter regions across diverse genes.

Dataset Structure

Total Variants: ~549,000
Species: Primates (Homo sapiens)
Variant Types: Insertions, deletions, and substitutions in promoter regions
Sequence Length: Up to 1,000bp promoter regions
MAF Range: < 0.001 (ultra rare)

Features

Basic Variant Information

gene: Gene symbol
species: Species name
clade: Evolutionary clade
chromosome: Chromosome
position: Genomic position
ref: Reference allele
alt: Alternative allele
variant_type: Type of variant (insertion/deletion/substitution)

Sequences

wt_sequence: Wild-type sequence
var_sequence: Variant sequence
sequence_length: Length of sequence
wt_sequence_start: Start position of sequence

Variant Annotations

filter: Variant filtering status
maf: Minor allele frequency

Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("cshearer/LOL-EVE-UltraRare")

# Access the data
print(f"Dataset size: {len(dataset['train'])}")
print(f"Features: {dataset['train'].features}")

# Example: Get variants by MAF threshold
rare_variants = dataset['train'].filter(lambda x: x['maf'] < 0.0001)
print(f"Ultra rare variants (MAF < 0.0001): {len(rare_variants)}")

# Example: Get variants by type
insertions = dataset['train'].filter(lambda x: x['variant_type'] == 'insertion')
deletions = dataset['train'].filter(lambda x: x['variant_type'] == 'deletion')
substitutions = dataset['train'].filter(lambda x: x['variant_type'] == 'substitution')

Citation

If you use this dataset in your research, please cite:

@article{loleve2024,
  title={LOL-EVE: Language of Life - Evolutionary Variant Effects},
  author={Your Name and Collaborators},
  journal={Nature},
  year={2024}
}

License

This dataset is released under the MIT License.