Employing machine learning for identifying antifungal compounds against Candida albicans

Abstract

Aims To evaluate the efficacy of a machine learning approach in developing classification and regression models for antifungal activity against Candida albicans. Materials & methods Utilized RF, SVM, and LightGBM algorithms to screen the eMolecules® library. Selected 17 virtual hits for in vitro assays. Results Eleven compounds showed activity against C. albicans. Compounds 1 and 17 inhibited C. albicans at 0.51 µM and 0.071 µM, respectively. Conclusions The RF model proved effective for virtual screening, demonstrating the success of the physicochemical classification and regression model in identifying new antifungal molecules against C. albicans. Plain Language Summary This study aimed to test a machine learning method for identifying substances that can combat Candida albicans, a type of fungus. The researchers utilized computer programs, including RF, SVM, and LightGBM, to analyze a large dataset of compounds. They selected 17 compounds for further testing in the lab. Of these, 11 compounds showed activity against C. albicans, and two compounds (1 and 17) could inhibit the fungus’ growth at very low concentrations. The results demonstrate that the machine learning model effectively identified new potential antifungal substances. This could aid in the development of more effective treatments for fungal infections.