High-Throughput Screening and Prediction Model Building for Novel Hemozoin Inhibitors Using Physicochemical Properties.

It is essential to continue the search for novel antimalarial drugs due to the current spread of resistance against artemisinin by Plasmodium falciparum parasites. In this study, we developed in silico models to predict hemozoin inhibitors as a potential first-step screening for novel antimalarials. An in vitro colorimetric high-throughput screening assay of hemozoin formation was used to identify hemozoin inhibitors from 9,600 structurally diverse compounds. The physicochemical properties of positive hits and randomly selected compounds were extracted from the ChemSpider database; they were used for developing prediction models to predict hemozoin inhibitors using two different approaches, i.e., traditional multivariate logistic regression and Bayesian model averaging. Our results showed that a total of 224 positive-hit compounds exhibited the ability to inhibit hemozoin formation, with 50% inhibitory concentrations (ICs) ranging from 3.1 μM to 199.5 μM. The best model according to traditional multivariate logistic regression included the three variables octanol-water partition coefficient, number of hydrogen bond donors, and number of atoms of hydrogen, while the best model according to Bayesian model averaging included the three variables octanol-water partition coefficient, number of hydrogen bond donors, and index of refraction. Both models had a good discriminatory power, with area under the curve values of 0.736 and 0.781 for the traditional multivariate model and Bayesian model averaging, respectively. In conclusion, the prediction models can be a new, useful, and cost-effective approach for the first screen of hemozoin inhibition-based antimalarial drug discovery.