Figure 1.

Comparison of the normalized wpc-index (0 for (worst) and 1 for (best)) between different ensemble approaches. The wpc-index is a composite score based on the ranking of the 18 test cancer cell lines with respect to their sensitivity to all the drug compounds and relative to the 35 training cancer cell lines. The rankings are based on the sensitivity of the cancer cell lines to each drug compound (from most to least effective). Ensemble of methods (1) corresponds to an ensemble model of random forest, SVM, and linear regression and where the predictions are made with the best-performing ensemble model of SVM. Ensemble of methods (2) corresponds to an ensemble of SVM with a radial kernel, SVM with a polynomial kernel, random forest, ridge regression, and boosting and where each method is applied on the five datasets and the results are then combined. Ensemble of methods (3) corresponds to an ensemble of k-nearest neighbor regression, linear least-squares regression, support vector regression, and gradient boosting regression and where the predictions are made with the best-performing ensemble model. SVM, support vector machine.