Add metadata.json

metadata.json

@@ -0,0 +1,14 @@
+{
+  "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"
+  ]
+}