A comparative study of support vector machine, artificial neural network and bayesian classifier for mutagenicity...

Area of Science:

• Computational chemistry
• Toxicology
• Bioinformatics

Background:

• Mutagenicity, the ability of chemicals to cause genetic mutations, poses significant risks in drug development.
• Early identification of mutagenicity is vital to prevent costly late-stage drug failures.

Purpose of the Study:

• To develop and compare computational models for predicting chemical mutagenicity.
• To assess the accuracy of Support Vector Machine (SVM), Artificial Neural Network (ANN), and Bayesian classifiers for mutagenicity prediction.

Main Methods:

• Utilized seventeen molecular descriptors to train and evaluate three distinct classification models.
• Implemented Radial Basis Function (RBF) kernel for the SVM classifier.
• Compared prediction accuracies of SVM, ANN, and Bayesian classifiers.

Main Results:

• The SVM classifier achieved the highest prediction accuracy at 71.73%.
• ANN classifier accuracy was 59.72%, and the Bayesian classifier accuracy was 66.61%.
• The RBF kernel-based SVM demonstrated superior performance in classifying mutagens and non-mutagens.

Conclusions:

• SVM-based models offer a more accurate approach for predicting chemical mutagenicity compared to ANN and Bayesian methods for the dataset studied.
• Early mutagenicity prediction using SVM can aid in de-risking drug development pipelines.