Performance Analysis of Drug Toxicity Prediction Using Machine Learning Approaches

Prediction of toxicity is a great challenge over the years. The development of a new drug or chemical compound is quite an expensive and complex process. Artificial intelligence and machine learning techniques give new insights into drug toxicity prediction. A quantitative structure-activity relationship has accomplished significant advancements in the field of toxicity prediction. This work presents a comparative analysis of nine machine-learning algorithms. The models applied are Gaussian Process, Linear Regression, Artificial Neural Network, SMO, Kstar, Bagging, Decision Tree, Random Forest, and Random Tree to predict toxicity. Feature selection algorithms in machine learning help to assign ranks to selected features depending upon the priorities and importance. This paper analyzes and identifies prominent features using feature selection analysis. The ranking is assigned to the features based on the weightage assigned. Ten folds of cross-validation are used to create a robust model. Our results provide the comparison of models and show the best prediction algorithms as Kstar, Random forest, and Bagging in terms of state of art parameters. Accuracy, coefficient of correlation, coefficient of determination, mean absolute error, and root mean squared error are evaluated to help build the optimized prediction model for toxicity prediction.