bgc-setnet / source /data /manifests /legacy_audit.json
whiteh4t's picture
Release final BGC retrieval checkpoints and model card
c87881a verified
Raw
History Blame Contribute Delete
1.18 kB
{
"atlas": {
"eligible_groups_at_least_5": 112,
"labeled_bgcs": 5476,
"labels_matching_mibig_accession": 5476,
"labels_with_at_least_3_direct_target_hits": 5476,
"reference_ids": 248,
"rows": 6953,
"unique_bgcs": 6953,
"unique_genomes": 182
},
"embeddings": {
"keys": 294604,
"missing_unique_gene_ids": 0,
"unique_gene_coverage": 1.0
},
"genes": {
"duplicate_composite_keys": 21652,
"duplicate_gene_ids_across_rows": 21653,
"rows": 316257,
"unique_gene_ids": 294604
},
"leakage": {
"finding": "Legacy labels are MIBiG-reference pseudo-labels and label-generating alignment fields must not be model inputs.",
"forbidden_atlas_feature_columns_present": [
"Alkaloid",
"NRP",
"Other",
"Polyketide",
"RiPP",
"Saccharide",
"Terpene",
"antibacterial",
"antifungal",
"avg_mibig_identity",
"cytotoxic",
"deepbgc_score",
"inhibitor",
"product_activity",
"product_class"
],
"label_generation_columns_present": [
"evalue",
"mibig_id",
"pident",
"qcovs"
]
},
"schema_version": 1
}