Algorithms for Big Data Analysis in Biology and Medicine Fall , 2017 - 8 Lecture 13 : January 16 , 2018

The drug development pipeline is long, very costly and can often end in failure. While more money and data has been put into developing new drugs, the number of approved drugs has not increased. Adverse side-effects during clinical trials are one of the reasons for nonapproval of a drug and they are detected late in the development pipeline, making them very costly. Predicting side-effects in advance can greatly benefit the drug development process. The method developed uses canonical correlation analysis (CCA) to predict drug sideeffects from chemical structure [1]. An advantage of using CCA is that it is trained on combined side-effect information and predicts all side-effects together as opposed to other classifiers, for example SVM, which will give a binary prediction for each individual sideeffect.