Deep Multitask Learning-Driven Discovery of New Compounds Targeting <i>Leishmania infantum</i>

Visceral leishmaniasis caused by <i>Leishmania infantum</i> is a severe and often fatal disease prevalent in low- and middle-income countries. Existing treatments are hampered by toxicity, high costs, and the emergence of drug resistance, highlighting the urgent need for novel therapeutics. In this context, we developed an explainable multitask learning (MTL) pipeline to predict the antileishmanial activity of compounds against three <i>Leishmania</i> species, with a primary focus on <i>L. infantum</i>. Then, we screened ∼1.3 million compounds from the ChemBridge database by using these models. This approach identified 20 putative hits, with nine compounds demonstrating significant <i>in vitro</i> antileishmanial activity against <i>L. infantum</i>. Three compounds exhibited notable potencies (IC₅₀ of 1.05-15.6 μM) and moderate cytotoxicities (CC₅₀ of 32.4 to >175 μM), positioning them as promising candidates for further hit-to-lead optimization. Our study underscores the effectiveness of multitask learning models in virtual screening, enabling the discovery of potent and selective antileishmanial compounds targeting <i>L. infantum</i>. Incorporating explainable techniques offers critical insights into the structural determinants of biological activity, aiding in the rational design and optimization of new therapeutics. These findings advocate for the potential of multitask learning methodologies to enhance hit rates in drug discovery for neglected tropical diseases.