Classification models for predicting the antimalarial activity against Plasmodium falciparum

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Predicting antimalarial activity is crucial for developing new treatments against Plasmodium falciparum.
• Machine learning offers promising approaches for quantitative structure-activity relationship (QSAR) studies.

Purpose of the Study:

• To develop and evaluate classification models for predicting antimalarial activity.
• To compare the performance of Support Vector Machine (SVM) and General Regression Neural Network (GRNN) models.

Main Methods:

• Utilized 15 molecular descriptors to build classification models for 4750 compounds.
• Trained and tested SVM and GRNN models on distinct datasets (3887 training, 863 test).
• Compared model performance against binary logistic regression (BLR) and previous classification models.

Main Results:

• SVM model achieved 89.5% (training) and 87.3% (test) prediction accuracy.
• GRNN model achieved 99.7% (training) and 88.9% (test) prediction accuracy.
• Both SVM and GRNN models outperformed BLR analysis.

Conclusions:

• SVM and GRNN models demonstrate satisfactory predictive ability for antimalarial activity.
• These models effectively predict activity using a reduced set of molecular descriptors.
• The developed models show potential for accelerating antimalarial drug discovery.