{
  "READY": true,
  "name": "vep_pathogenic_coding",
  "task": "binary",
  "num_classes": 2,
  "cat_idx": [
    0,
    1,
    2,
    3,
    4,
    5,
    6,
    7,
    8,
    9,
    10,
    11,
    12,
    13,
    14,
    15,
    16,
    17,
    18,
    19,
    20,
    21,
    22,
    23,
    24,
    25,
    26,
    27,
    28,
    29,
    30,
    31,
    32,
    33,
    34,
    35,
    36,
    37,
    38,
    39,
    40,
    41,
    42,
    43,
    44,
    45,
    46,
    47,
    48,
    49,
    50,
    51,
    52,
    53,
    54,
    55,
    56,
    57,
    58,
    59,
    60,
    61,
    62,
    63,
    64,
    65
  ],
  "num_samples": 37166,
  "num_features": 98,
  "imbalance_ratio": 1.1365,
  "description": "Classify coding single-nucleotide variants as pathogenic (ClinVar, label 1) or common (gnomAD MAF > 5%, label 0) from DNA sequence context only. Variants are strand-canonicalised (reverse-complemented when the reference allele is a purine, so it is always C or T). Features: GRCh38 reference bases at offsets -32..+32 around the variant (the central base is the reference allele) and the alternative allele (66 categorical), plus canonical 3-mer frequencies (k-mer and reverse complement merged) of the 513-bp reference window centred on the variant (32 numeric). Allele-frequency fields, annotation fields (VEP consequence, ClinVar IDs/review status) and genomic coordinates are excluded because they define or reveal the label. chrX/chrY variants (all pathogenic by construction) and two variants listed with both labels are removed. Split by whole chromosomes: test = chr1, chr4, chr21; validation = chr5, chr6, chr11; train = remaining autosomes.",
  "source": "InstaDeepAI/genomics-long-range-benchmark (Hugging Face), task variant_effect_pathogenic (ClinVar, coding SNVs): ClinVar pathogenic variants vs gnomAD v3.1.2 common variants (MAF > 5%, AN >= 25,000) (Kao et al., 2024). Sequence context from the UCSC hg38 (GRCh38) reference genome.",
  "label": "genomics"
}