PCA-mapping and machine learning identify multiple chemotypes with antiplasmodial, β-haematin and haemozoin formation...

Inhibiting the formation of haemozoin remains a robust, non-genomic target for antiplasmodial drug discovery. While current efforts focus on multistage active compounds that are not pure haemozoin formation inhibitors, this class offers advantages including selectivity for Plasmodium and is associated with lowered resistance risks. With a view to developing cost-effective tools to streamline the discovery process, we developed a high-throughput virtual screening (HTVS) workflow integrating both ligand- and structure-based techniques. Here, we employed a principal component analysis (PCA) enrichment map to prioritise a commercial library for potential β-haematin (βH, the synthetic analogue of haemozoin) formation inhibitors. Compounds identified within the area of enrichment were subsequently evaluated using structure-based molecular docking to the βH crystal. A support vector machine (SVM) model was then applied to predict antiplasmodial activity, after which a diverse subset of the highest-ranking hits was validated through detergent-mediated βH formation inhibition and whole-cell Plasmodium falciparum (Pf) assays. Three frontrunner compounds (1, 2, and 14) were found to possess potent, sub-micromolar Pf inhibitory activity. Furthermore, compounds 1 and 2 were found to possess βH formation inhibitory activity, analogous to chloroquine, with cellular fractionation confirming haemozoin formation inhibition as a contributor to the whole-cell mechanism of action of compound 1. This accessible and efficient HTVS workflow provides a successful framework for identifying new chemotypes targeting the haem detoxification pathway and, through PCA modification, potentially other essential parasite processes.