AutoFilter: A Low-Cost Biocomputational Framework for High-Throughput Screening of Chemical Databases and Identification of Novel Malaria Inhibitors Targeting Plasmodium Falciparum

Abstract

Malaria is the third deadliest disease, with approximately 249 million cases annually, particularly in tropical regions. Caused by Plasmodium parasites transmitted through the bite of Anopheles mosquitoes, malaria remains a significant global health burden and is increasingly difficult to treat due to rising drug resistance. Drug discovery for malaria is both costly and time- consuming, typically requiring over a decade and around $3 billion before a compound gains approval. To address this challenge, AutoFilter was developed: a low-cost and novel biocompu- tational framework that integrates machine learning (ML) and screening tools to streamline the filtering of large chemical databases for more efficient drug discovery. AutoFilter sequentially screens compounds that violate basic chemical filters such as Lipinski’s Rule of 5 (Lipinski et al., 1997), Veber’s (Veber et al., 2002), and PAINS; performs molecular docking and analyzes post- docking interactions; conducts ADME filtration to identify compounds with favorable drug-like properties; employs an ML model to predict toxicity and synthetic accessibility; and finally applies molecular dynamics (MD) simulations to refine compound stability. AutoFilter was applied to screen the ChEMBL database, which contains 2.4 million bioactive compounds, to identify malaria inhibitors targeting Plasmodium falciparum apPOL. The five selected compounds demonstrated high inhibition performance and favorable drug-like properties and are currently undergoing in vitro trials. As the first integrated biocomputational framework for chemical database screening, AutoFilter is a transformative tool for drug discovery across diverse diseases, efficiently identifying inhibitors while reducing costs and time by 50%, with the profound potential to save lives worldwide.