Generative AI Meets Natural Product Chemistry: A Critical Review of Molecular Design, Synthetic Feasibility, Bioactivity Prediction, and Safety Validation

Download PDF

Generative AI is increasingly used to support molecular design, chemical space exploration, and compound prioritisation, creating new opportunities for natural product chemistry while also raising important questions about chemical realism and validation. This critical review evaluates how generative models may contribute to natural product-inspired scaffold generation, analogue design, molecular novelty, diversity, bioactivity prediction, synthetic feasibility assessment, ADMET screening, toxicity prediction, and safety validation. The review takes a cautious interpretive approach, treating generated structures as design hypotheses rather than validated chemical, biological, or translational entities. Particular attention is given to the structural complexity of natural products, including stereochemistry, biosynthetic diversity, functional-group density, and the difficulty of distinguishing natural product likeness from biological relevance. The synthesis argues that generative AI can expand ideation, prioritise candidate chemical regions, and integrate multi-objective design constraints, but its outputs remain vulnerable to invalid structures, unrealistic stereochemistry, synthetic accessibility overestimation, false-positive bioactivity prediction, target-context mismatch, ADMET uncertainty, toxicity underprediction, benchmark bias, data leakage, and weak uncertainty reporting. The main conclusion is that generative AI has genuine value for natural product-inspired molecular exploration only when coupled to chemically informed constraints, retrosynthetic review, experimentally grounded bioactivity testing, safety assessment, uncertainty calibration, and transparent reporting. Future research should prioritise domain-specific datasets, stereochemistry-aware generation, synthesis-linked evaluation, biologically meaningful objectives, and validation pipelines that preserve the expertise of natural product chemists.