Figure 1.

Dataset and model description. Considering the drug response data as learning data for our prediction target (ln[IC₅₀] values), we combined two types of data, including genomic information (gene expression and mutation profiles). This yielded expression (EC-11K) and mutation (MC-9K) datasets for the drug-response prediction model. We set two input settings to construct drug response, prediction models. Settings 1 and 2 handle gene expression profiles (mutation profiles for setting 2) to predict ln(IC₅₀) values for an individual drug in one model, such that settings 1 and 2 had a total of 24 models (for prediction of drug response for 24 drugs). We have used three abbreviations: E (expression), M (mutation), and C (drug response of CCLE cell lines, ln[IC₅₀]). CCLE, Cancer Cell Line Encyclopedia.