NaturalProductMech

3136 natural product structures, one record per Standard InChIKey, harmonized from 10 adopted sources.

A producer is not an occurrence. This corpus keeps the two apart because most sources do not. 3441 producer claims say a taxon makes the compound, and only 840 of those rest on evidence that addressed the organism. 2351 records carry a cited occurrence, which says only that someone found the compound there. Nothing promotes one to the other.

Explore maps & browse

Faceted browse

Filter 3136 compounds by pathway, causal producer, gene cluster, or target

Structure map (2D UMAP)

Chiral Morgan fingerprint Tanimoto neighborhoods

Corpus map (text PaCMAP)

2D projection of curated mechanism & producer descriptions

Sequence map (pLM)

2D PaCMAP of ESM-2 embeddings for biosynthetic enzymes & targets

By filing pathway

The pathway is NPClassifier's, so it is computed, pinned per record and versioned. The asserted classifications live beside it on each record.

Alkaloids

512 records

Amino acids and peptides

661 records

Carbohydrates

104 records

Fatty acids

87 records

Polyketides

869 records

Shikimates and phenylpropanoids

64 records

Terpenoids

213 records

Unclassified

626 records

What the corpus holds

Field	Records carrying at least one
biosynthesis_origin	3136
producer_taxon_groups	3097
occurrence_taxon_groups	2351
producer_organisms	3097
occurrences	2351
biosynthetic_gene_clusters	3123
biosynthetic_pathway	216
bioactivities	174
bioactivity_summary	314
molecular_targets	35
causal_graphs	317
related_records	217
discussions	362

Downloads & Schema

Machine-readable data exports, LinkML schema definitions, and embedding coordinate tables.

LinkML Schema (YAML)

Formal schema for natural product structures, producer claims, occurrences, BGCs, and causal graphs (mech_shared.yaml, history.yaml)

Curated YAML Corpus (3136 records)

One YAML file per Standard InChIKey organized by NPClassifier pathway

Structure Map Coordinates (JSON)

2D chiral Morgan fingerprint UMAP coordinates and k-NN neighborhoods (chemical-search.json)

Protein Sequence Map (JSON)

ESM-2 protein language model PaCMAP coordinates for biosynthetic enzymes & molecular targets

Adopted Sources & Manifest

Source provenance registry and data/raw/MANIFEST.yaml

Generated from sources retrieved 2026-10-31 (3136 records · from sources retrieved 2026-10-31 · corpus SHA-256 2838a2004e52).