Speaker
Description
Chagas disease, caused by Trypanosoma cruzi, affects an estimated 6–7 million people globally, predominantly in low‑income regions of Latin America, sub‑Saharan Africa, and Southeast Asia. Despite its significant health burden, drug discovery for Chagas disease remains neglected, with current treatments being outdated, of limited efficacy, and often associated with severe side effects. This study sought to identify potential T. cruzi inhibitors through an integrated computational approach combining quantitative structure–activity relationship (QSAR) modeling and molecular docking. Experimental datasets were sourced from Google Scholar and PubChem, curated, and modeled using Spartan 14. Molecular descriptors were generated with Rowan Descriptors and analyzed in Python (scikit‑learn) using correlation filtering and feature selection. Redundant descriptors (correlation ≥ 0.8) were removed, reducing 2,145 descriptors to 22. Six supervised machine learning models were trained, with Support Vector Regression (SVR) initially performing best (R² = 0.116). Refinement reduced the feature set to nine descriptors, improving SVR performance substantially (R² = 0.67). Linear regression models guided by Topliss’s rule generated interpretable equations, which informed the design of new compounds based on descriptor influence on biological activity. Molecular docking was carried out against three essential T. cruzi proteins: sterol 14-α-demethylase (CYP51), trypanothione reductase (TR), and cysteine synthase. The docking results highlight LM4 and LM6 as the most promising inhibitors against T. cruzi. LM4 benefits from dual-target inhibition, while LM6 exploits a parasite-selective pathway in cysteine biosynthesis. The interaction strength analysis suggests that both compounds are capable of forming stable hydrogen bonds and hydrophobic contacts within the catalytic sites of their targets, with LM6 showing the strongest predicted affinity.
Keywords: Chagas Disease, Trypanosoma cruzi, QSAR, Molecular Descriptors, Supervised Machine Learning, Molecular Docking.