Mapping Natural Product Data Ecosystems for Computational Pharmacology: Phytochemical Libraries, Bioassays, Targets, Pathways, Omics Data, and Clinical Signals

Download PDF

Natural-product computational pharmacology increasingly depends on heterogeneous data that extend from biological source identity and phytochemical composition to bioactivity, molecular targets, pathways, omics states, pharmacokinetics, safety observations, and clinical signals. Yet these data are generated under different experimental conditions, represented through incompatible identifiers and schemas, and supported by evidence of unequal maturity. This Original Taxonomy Article maps that ecosystem and proposes a multi-dimensional classification framework designed specifically for natural-product computational pharmacology. The analysis distinguishes phytochemical records from experimentally supported occurrence and quantitative composition, structure-resolved identities from names and synonyms, and spectral or computational annotations from confirmed structures. Bioassay data are classified according to assay type, endpoint, concentration context, metadata completeness, and evidential status. Target evidence is separated into measured, functional, curated, predicted, and text-derived relations, while network and pathway associations are distinguished from direct target engagement. Transcriptomic, proteomic, metabolomic, multi-omics, single-cell, and spatial data are treated as context-dependent molecular-state evidence rather than automatic mechanistic confirmation. Pharmacokinetic, toxicity, adverse-event, and clinical data are further separated according to exposure context, provenance, causal limitations, and translational proximity. The proposed taxonomy organizes data through nine top-level classes and cross-cutting dimensions of evidence mode, biological scale, computational role, provenance, evidence maturity, interoperability readiness, uncertainty, validation status, and translational proximity. It further identifies interoperability gates involving identity resolution, metadata harmonization, semantic alignment, version control, and uncertainty preservation. The framework is intended to support more auditable databases, knowledge graphs, multimodal models, candidate-prioritization systems, and translation-aware computational platforms without assuming that data aggregation removes upstream uncertainty or makes heterogeneous evidence types equivalent.