In this chapter, a new method to evaluate the reliability of predicting new uses of existing drugs was proposed. The prediction was performed with a support vector machine (SVM) using various data. Because the reliability of prediction could not be evaluated based on the output of an SVM, which was binary, the proposed method evaluated the reliability as a product of a distance from the separating hyperplane of the SVM and a similarity between the disease targeted by the drug and a candidate disease. A validation using real data revealed that the performance of the proposed method was promising.
Keywords: Chemical structure; Drug repositioning; Drug target; Machine learning; Reliability score; Side effect; Support vector machine (SVM).