Supporting Information for Article: A copula based approach for design of multivariate random forests for drug sensitivity prediction.

RF, VMRF, CMRF results (5 fold cross validation) with and without prior feature selection (Table A). Results for CCLE Dataset drug sensitivity prediction for a drugset with 4 drugs in the form of correlation coefficients for RF, VMRF, CMRF and KBMTL approaches (Table B). Results for GDSC Dataset drug sensitivity prediction for a drugset with 3 drugs in the form of correlation coefficients for RF, VMRF, CMRF and KBMTL approaches (Table C). Results for GDSC Dataset drug sensitivity prediction for a drugset with 140 drugs in the form of correlation coefficients is shown (only 15 drugs that are common with CCLE are shown in detail while the average represents the average of all 140 drugs) (Table D). Results for GDSC Dataset drug sensitivity prediction for a drugset with 140 drugs in the form of NRMSE is shown (only 15 drugs that are common with CCLE are shown in detail while the average represents the average of all 140 drugs) (Table E). Results for CCLE Dataset drug sensitivity prediction for the combined set of 24 drugs in the form of correlation coefficients (Table F). Results for CCLE Dataset drug sensitivity prediction for the combined set of 24 drugs in the form of Normalized Root Mean Square Error (Table G). Comparison of α for different sets of synthetic data with and without noise added to the drug response (Table H). Comparison of α for different amount of random subset of the original samples in a specific synthetic data. Original number of samples were 350 in this specific example (Table I). Simulation time for different drugsets in GDSC data. The reported simulation times are the time needed to generate complete result for all drugs in a drug set for 5 fold cross validation (Table J). Simulation time for different drugsets in GDSC data. The reported simulation times are the time needed to generate complete result for all drugs in a drug set for 30–70 case (Table K). Simulation time for different methods for all drugs of GDSC dataset (140) and CCLE dataset (24). The reported simulation times are the time (in seconds) needed to generate complete result for all drugs for 30–70 case (Table L). (PDF)