An Integrative Analysis of Anti-Cancer Drug Response

In this project, we perform an integrative analysis of 24 kinds of anti-cancer drug response on around 400 cancer patients. First, we use Lasso to build a prediction system for each individual drug response. Then a graph Lasso is employed to utilize the correlation between multiple tasks in order to build a more robust prediction system. Finally, we further perform supervised and unsupervised categorization of patients into multiple cancer types. Our analysis shows gene expression is most discriminative and useful in cancer categorization and drug response prediction.