QSAR based predictive modeling for anti-malarial molecules.

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Parasitology

Background:

• Malaria remains a major global health threat, particularly in Africa, with increasing drug resistance necessitating novel therapeutic strategies.
• The apicoplast, an organelle unique to Plasmodium parasites and absent in humans, represents a promising target for antimalarial drug development.
• Machine learning (ML) offers powerful tools for accelerating drug discovery through the analysis of large chemical datasets.

Purpose of the Study:

• To establish a standardized protocol for developing Quantitative Structure-Activity Relationship (QSAR) models for antimalarial drug discovery.
• To identify effective ML methods for predicting compound activity against apicoplast targets.
• To facilitate the screening of extensive chemical libraries for novel antimalarial drug candidates.

Main Methods:

• Development of a protocol for building molecular descriptor-based predictive models (QSAR models).
• Utilized training data from apicoplast-specific bioassays.
• Employed and evaluated multiple machine learning algorithms: Generalized Linear Models (GLM), Random Forest (RF), C5.0, Support Vector Machines (SVM), K-Nearest Neighbour, and Naive Bayes.

Main Results:

• The C5.0, SVM, and RF models demonstrated superior performance compared to other methods evaluated.
• These top-performing models achieved comparable accuracy on the test dataset.
• The developed protocol provides a robust framework for QSAR model generation.

Conclusions:

• The established protocol effectively generates predictive QSAR models for antimalarial drug discovery targeting the apicoplast.
• Machine learning, particularly C5.0, SVM, and RF, are highly effective for screening potential antimalarial compounds.
• This approach aids in the identification of new drug leads to combat drug-resistant malaria.