Machine Learning-Based QSAR Models for Discovery of Inhibitors Targeting Leishmania infantum Amastigotes

Area of Science:

• Computational chemistry and pharmacology
• Infectious diseases and tropical medicine
• Drug discovery and development

Background:

• Leishmaniasis, a neglected tropical disease caused by Leishmania parasites, presents challenges in treatment, particularly in young children.
• Leishmania infantum infections are increasing, necessitating novel therapeutic strategies.
• In silico methods offer a promising avenue for accelerating the identification of new antileishmanial drugs.

Purpose of the Study:

• To identify novel compounds with potential activity against Leishmania infantum amastigotes.
• To leverage artificial intelligence (AI)-based classification models for antileishmanial drug discovery.

Main Methods:

• Construction of a curated database of compounds with known biological activity.
• Utilizing molecular descriptors and unsupervised k-means clustering for dataset partitioning.
• Development and validation of supervised machine learning models (IBk, J48, MLP, SMO) on the WEKA platform.

Main Results:

• All developed AI models demonstrated high classification accuracy (>80%) on both training and external prediction sets.
• Validated models were employed for virtual screening of DrugBank and synthetic compound libraries.
• Identification of 120 compounds exhibiting potential activity against Leishmania infantum amastigotes.

Conclusions:

• AI-based Quantitative Structure-Activity Relationship (QSAR) models are effective for prioritizing antileishmanial drug candidates.
• The integration of molecular descriptors, machine learning, and virtual screening provides an efficient strategy for drug discovery.
• This study highlights the utility of AI in accelerating the search for treatments against neglected tropical diseases like leishmaniasis.