Implementation of Gravitational Search Algorithm - Ensemble in Predicting of Drug Side Effect: Case Study Hepatobiliary Disorders

A drug is a mixture of substances that can prevent, reduce and cure disease. Besides being able to prevent disease, drugs can cause side effects. It is the fourth leading cause of death in America and causes as many as 100,000 deaths each year. Many researchers identify drugs by combining compounds (receptors and enzymes), to produce predictions of drug side effects. But traditional experimentation and drug development are time-consuming and expensive. In vitro use is more difficult because biochemical tests must test cellular compounds, but many drugs target proteins that have not been described. In silico method is considered quite effective due to its ability to produce good predictions and new insights about how drugs work and the mechanism of side effects. In this study, a prediction model for drug side effects was developed using the Gravitational Search Algorithm (GSA) for feature selection and the ensemble method for building a prediction model with the aim of drug discovery in a case study of hepatobiliary disorders. with three methods, namely Random Forest, Adaptive Boosting (AdaBoost), and Extreme Gradient Boosting (XGBoost). The best model was obtained from Random Forest model with accuracy and F1 scores of 0.68 and 0.77, respectively.