Implementation of Firefly Algorithm-Support Vector Machine in Classifying Clinical Trial Toxicity

Drugs play a crucial role in improving human health and well-being. Before being approved for public use, drugs must undergo clinical trials to assess their toxicity and ensure safety. However, conventional methods, such as animal testing, are expensive, time-consuming, ethically controversial, and labor-intensive. To address these challenges, alternative approaches for toxicity prediction are essential. This study focuses on the Support Vector Machine (SVM), a supervised Machine Learning (ML) algorithm, which is highly effective for classification tasks. The kernels used in this study are linear, poly, and RBF. Additionally, the Firefly Algorithm (FA) is employed as an optimization technique for feature selection, aiming to improve model accuracy and reduce data dimensionality. The study explores the implementation of the FA-SVM hybrid approach to predict drug toxicity using a dataset containing SMILES representations, FDA approval and toxicity labels. The study compares the accuracy achieved by the different SVM kernels, with results indicating that the linear kernel achieved an F1-Scores of 0.980, the poly kernel 0.983, and the RBF kernel 0.983.