Malaria which is mainly caused by Plasmodium falciparum parasite remains a devastating public health concern, necessitating the need to develop new antimalarial agents. P. falciparum heat shock protein 90 (Hsp90), is indispensable for parasite survival and a promising drug target. Inhibitors targeting the ATP-binding pocket of the N-terminal domain have anti-Plasmodium effects. We proposed a de novo active learning (AL) driven method in tandem with docking to predict inhibitors with unique scaffolds and preferential selectivity towards PfHsp90. Reference compounds, predicted to bind PfHsp90 at the ATP-binding pocket and possessing anti-Plasmodium activities, were used to generate 10,000 unique derivatives and to build the Auto-quantitative structures activity relationships (QSAR) models. Glide docking was performed to predict the docking scores of the derivatives and > 15,000 compounds obtained from the ChEMBL database. Re-iterative training and testing of the models was performed until the optimum Kennel-based Partial Least Square (KPLS) regression model with a regression coefficient R2 = 0.75 for the training set and squared correlation prediction Q2 = 0.62 for the test set reached convergence. Rescoring using induced fit docking and molecular dynamics simulations enabled us to prioritize 15 ATP/ADP-like design ideas for purchase. The compounds exerted moderate activity towards P. falciparum NF54 strain with IC50 values of ≤ 6μM and displayed moderate to weak affinity towards PfHsp90 (KD range: 13.5-19.9μM) comparable to the reported affinity of ADP. The most potent compound was FTN-T5 (PfN54 IC50:1.44μM; HepG2/CHO cells SI≥ 29) which bound to PfHsp90 with moderate affinity (KD:7.7μM), providing a starting point for optimization efforts. Our work demonstrates the great utility of AL for the rapid identification of novel molecules for drug discovery (i.e., hit identification). The potency of FTN-T5 will be critical for designing species-selective inhibitors towards developing more efficient agents against malaria.
Read full abstract