Drug-Target Interaction Network Predictions for Drug Repurposing Using LASSO-Based Regularized Linear Classification Model

摘要

It has been well-known that biological and experimental methods for drug discovery are time-consuming and expensive. New efforts have been explored to perform drug repurposing through predicting drug-target interaction networks using biological and chemical properties of drugs and targets. However, due to the high-dimensional nature of the data sets extracted from drugs and targets, which have hundreds of thousands of features and relatively small numbers of samples, traditional machine learning approaches, such as logistic regression analysis, cannot analyze these data efficiently. To overcome this issue, we proposed a LASSO-based regularized linear classification model to predict drug-target interactions, which were used for drug repurposing for inflammatory bowel disease. Experiments showed that the model out performed the traditional logistic regression model.