Autoencoder Based Feature Selection Method for Classification of Anticancer Drug Response.

摘要: Anticancer drug responses can be varied for individual patients. This difference is mainly caused by genetic reasons, like mutations and RNA expression. Thus, these features are often used to construct classification models predict the response. research focuses on feature selection issue models. Because of vast dimensions space predicting response, autoencoder network was first built, a subset inputs with important contribution selected. Then using Boruta algorithm, further small set determined random forest, which Two datasets, GDSC CCLE, were illustrate efficiency proposed method.