Counter-Screens Before Rankings: A Falsification-First Strategy for Eliminating Artefactual, Promiscuous, and Mechanistically Implausible Natural Product Leads

Download PDF

Natural-product lead discovery increasingly combines biochemical screening, phenotypic assays, virtual screening, machine learning, and target-deconvolution technologies, yet candidate ranking can still reward positive signals before plausible alternative explanations have been actively challenged. This creates an evidential asymmetry: potency, docking score, predicted affinity, phenotypic magnitude, or apparent target breadth can raise a candidate's priority even when the signal could arise from assay interference, aggregation, reactive chemistry, impurities, nonspecific activity, or an implausible target assignment. This article develops an original decision framework that reverses that order. Rather than asking first which candidate ranks highest, it asks which candidate can survive structured attempts to disprove the interpretation attached to its activity. The analysis distinguishes artefact from promiscuity and promiscuity from purposeful polypharmacology, while treating mechanistic plausibility, target engagement, and orthogonal functional reproduction as separate evidentiary requirements. The proposed framework organizes counter-screens as sequential challenges to the lead hypothesis and reserves final ranking for candidates that remain interpretable after those challenges. Importantly, failure does not always mandate rejection: discordant evidence can identify an impurity, alternative target, context-sensitive mechanism, or unresolved assay limitation that warrants investigation. The framework is a decision architecture rather than a validated universal screening algorithm. Its intended value is to make negative evidence explicit, preserve uncertainty, and prevent attractive scores from acquiring more evidential meaning than the experiments support. Prospective, multi-campaign testing is required before claims of improved attrition, efficiency, or translational success are justified.