Replacing Single-Point Potency with Dose–Time Response Surfaces for Computational Prioritization of Natural Product Candidates

Download PDF

Natural-product discovery commonly compresses pharmacological activity into potency values obtained at selected concentrations and observation times. Such summaries are useful for screening, but they can conceal whether activity is rapid or delayed, transient or persistent, growth-suppressive or lethal, and whether apparent potency changes as the biological system evolves. This article develops an original pharmacodynamic framework in which candidate response is represented as a dose–time surface rather than a single potency estimate. The framework treats concentration and observation time as explicit coordinates, preserves uncertainty and assay context, and separates scalar feature extraction from direct learning of the full response landscape. It further proposes that candidate prioritization should depend on the therapeutic objective: rapid suppression, durable effect, recoverability, maximal effect, or exposure-compatible response may imply different preferred regions of the same surface. Dynamic phenotypes are interpreted descriptively unless orthogonal evidence establishes mechanism, and in vitro surface geometry is not equated with target engagement, clinical exposure, or therapeutic efficacy. The framework also identifies design requirements for surface construction, including longitudinal or repeated measurements, adequate concentration coverage, replication, normalization, and uncertainty-aware interpolation. Its central contribution is a proposed decision architecture for comparing candidates whose apparent potency rankings diverge across time. The approach remains conceptual and requires prospective testing against simpler ranking strategies, with held-out compounds and biological contexts, before any claim of improved predictive or translational performance is justified.