Mal-Predict: Machine Learning-Guided Rapid Virtual Screening of Compounds Against Selected Targets of Plasmodium Falciparum Validated Using Molecular Dynamics Simulation

Why It Matters

By automating the triage of millions of molecules, AI‑driven virtual screening can dramatically shorten the antimalarial lead‑identification timeline and cut costly wet‑lab experiments.

Key Takeaways

• RF model achieved 0.912 AUC on antimalarial data
• Screened 1.9 M compounds across three chemical libraries
• EN52 and NP83 outperformed reference ligands in binding energy
• Mal‑Predict merges ML classification with docking and MD validation

Pulse Analysis

Malaria remains a global health priority, yet the pipeline for new antimalarial agents is hampered by high attrition rates and lengthy discovery cycles. Traditional high‑throughput screening can test only a fraction of chemical space, leaving many promising scaffolds unexplored. Recent advances in artificial intelligence, particularly in supervised learning, have opened pathways to evaluate billions of molecules in silico, offering a cost‑effective complement to experimental assays. By leveraging large public datasets such as ChEMBL and PubChem, researchers can train robust classifiers that discern subtle structure‑activity relationships specific to Plasmodium falciparum targets.

The Mal‑Predict platform exemplifies this new paradigm. Using a Random Forest classifier that achieved an impressive 0.912 area‑under‑the‑curve, the team rapidly labeled 1.9 million compounds from DrugBank, natural‑product collections, and the Enamine‑Real library as likely actives or inactives. High‑confidence hits were then subjected to molecular docking against priority PfPKG and F/GGPPS proteins, followed by molecular dynamics simulations and MM‑GBSA free‑energy calculations. Compounds EN52 and NP83 emerged with binding energies of ‑48.05 ± 3.91 and ‑52.67 ± 4.43 kcal/mol, respectively, outperforming the co‑crystallized ligands and indicating strong thermodynamic stability within the active sites.

The integration of machine learning, docking, and dynamics in a single pipeline offers tangible benefits for pharmaceutical firms and academic labs alike. It reduces the experimental burden by focusing resources on a handful of high‑potential candidates, accelerates the transition from hit identification to lead optimization, and provides a reproducible framework that can be adapted to other parasitic diseases. As the industry embraces AI‑augmented drug discovery, tools like Mal‑Predict are poised to become standard components of early‑stage pipelines, driving faster, cheaper, and more innovative antimalarial therapeutics.