{
  "READY": true,
  "name": "true_melting_point",
  "task": "regression",
  "cat_idx": [],
  "num_samples": 12144,
  "num_features": 480,
  "description": "Sequence-level regression task predicting the melting temperature of proteins. Data originates from the Thermostability Atlas, compiled via a mass spectrometry-based proteomic approach. Splits follow the 'mixed' strategy from FLIP: sequences are clustered at 20% identity, with 80% of clusters assigned to train and 20% to test, avoiding over-emphasis of large clusters. The source provides coding DNA sequences (CDS); they were translated with the standard genetic code before featurization, and nonsense mutants (CDS with a premature in-frame stop codon) were removed from all splits. Features: mean-pooled ESM-2 (esm2_t12_35M_UR50D) embeddings of the translated protein.",
  "source": "InstaDeepAI/true-cds-protein-tasks (Hugging Face). Original data from the Thermostability Atlas (mass spectrometry-based proteomics). Splits follow the 'mixed' split strategy from FLIP: sequences clustered at 20% identity, 80% of clusters assigned to train and 20% to test.",
  "label": "Proteomics",
  "sub_labels": [
    "Protein sequence",
    "Translated CDS"
  ]
}