Advancing antimalarial drug discovery: ensemble machine learning models for predicting PfPK6 inhibitor activity

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Machine Learning

Background:

• Malaria remains a major global health issue, with increasing drug resistance necessitating novel therapeutic strategies.
• Targeting Plasmodium falciparum protein kinase 6 (PfPK6) is a promising avenue for antimalarial drug development.

Purpose of the Study:

• To develop robust predictive models for identifying novel Plasmodium falciparum protein kinase 6 (PfPK6) inhibitors.
• To leverage machine learning techniques for the rational design of new antimalarial agents.

Main Methods:

• Utilized machine learning algorithms including Random Forest, Relevance Vector Machine, Support Vector Machine, Cubist, Artificial Neural Networks, and XGBoost for predictive modeling.
• Employed Classification and Regression Trees (CART) for molecular descriptor refinement and identified key features.
• Developed ensemble regression and classification models, validated through applicability domain analysis and y-randomization tests.

Main Results:

• The consensus regression model achieved high predictive performance (R²Test=0.94, Q²CV=0.90), outperforming individual models.
• The ensemble classification model demonstrated strong accuracy (91%) and sensitivity (93%).
• Applicability domain analysis confirmed model robustness with 96% coverage.

Conclusions:

• Ensemble machine learning approaches are highly effective for predicting antimalarial drug efficacy.
• The developed models provide valuable insights for the rational design of novel PfPK6 inhibitors to combat malaria.