THEOBROMA v1.34: an open multi-kingdom natural-products database

THEOBROMA is an aggregated open natural-products database containing 1,133,004 compounds from 29 source databases across six biological kingdoms. The corpus preserves full-stereochemistry InChIKeys and exposes the resulting 486,215 stereoisomer families as a directly queryable resource. Provenance is tracked at compound granularity for both classification (curated source / high-confidence inferred / exploratory inferred) and licensing (per-compound source license, with corpus-level aggregation). Open (CC BY 4.0 + CC0) compound coverage is 88.8% of the corpus. License tiers are recorded per-compound in the license_tier column for downstream filtering. The deposit comprises six relational CSV tables (compounds, compound_synonyms, resolved_taxonomy, admet, scaffolds, compound_taxonomy), precomputed ChemBERTa embeddings and a FAISS HNSW similarity index, fingerprint matrices, classification artifacts, and an Optuna hyperparameter study. The accompanying live database is at https://theobroma.l3s.uni-hannover.de and source code at https://github.com/ThorKlm/theobroma. This deposit accompanies a manuscript in preparation; the preprint will be posted on bioRxiv.