Molecular Docking, Molecular Dynamics, and Machine Learning in Phytopharmacology: An Umbrella Review of Computational Evidence for Natural Product Therapeutics

Download PDF

Phytopharmacology and natural product therapeutics remain important areas of drug discovery because medicinal plants, phytochemicals, marine metabolites, microbial products, and natural product scaffolds continue to provide biologically meaningful chemical diversity. Computational methods are increasingly used to prioritize this chemical space, but the interpretive strength of in silico evidence varies across molecular docking, molecular dynamics, machine learning, network pharmacology, ADMET prediction, toxicity screening, and virtual screening workflows. This umbrella review synthesizes review-level evidence on how these computational approaches are used to support natural product therapeutic prioritization while distinguishing computational plausibility from pharmacological proof. The review focuses on molecular docking as a method for ligand–target prioritization, binding-site hypothesis generation, and virtual screening; molecular dynamics as a method for examining interaction stability, conformational behavior, and post-docking plausibility; and machine learning as a method for bioactivity prediction, target prediction, ADMET modeling, toxicity prediction, QSAR, database mining, and lead prioritization. Cross-method synthesis is used to evaluate whether convergent computational evidence strengthens hypothesis generation or risks reinforcing uncertainty when methods share biased inputs, poorly curated structures, weak target assumptions, or limited validation. The review identifies recurring gaps in methodological reporting, reproducibility, dataset quality, applicability-domain assessment, experimental confirmation, and translational interpretation. Overall, computational phytopharmacology can support rational prioritization and mechanism-oriented hypothesis generation, but therapeutic claims require reproducible workflows, verified chemical structures, transparent evidence synthesis, pharmacological validation, and translationally relevant experimental confirmation.