metadata.json

{
  "READY": true,
  "name": "true_betalactamase_complete",
  "task": "regression",
  "cat_idx": [],
  "num_samples": 12604,
  "num_features": 480,
  "description": "Sequence-level regression task predicting the fitness of all single codon substitutions in the TEM-1 Beta-lactamase gene. Labels measure the ability to confer ampicillin resistance. The Complete dataset contains all coding sequences (CDS) except those degenerate with any test-set CDS, using raw CDS fitness values. Train and test sets are derived from Firnberg et al. Features are mean-pooled ESM-2 (esm2_t12_35M_UR50D) sequence embeddings.",
  "source": "InstaDeepAI/true-cds-protein-tasks (Hugging Face). Original data from Firnberg et al. \u2014 systematic fitness landscape of all single codon mutations in the TEM-1 Beta-lactamase gene synthesized in E. coli. Fitness measured as a weighted average of allele counts over ampicillin-concentration plates, normalised by WT fitness.",
  "label": "Proteomics",
  "sub_labels": [
    "CDS-seq"
  ]
}