Figure 3.

Subchallenge 1 feature selection and learning through an ensemble model. (a) Feature level—from the training set that contains five different types of data, multiple training datasets are created (CNV, RPPA, methylation, gene expression, and RNA-seq). (b) Classifier level—each of the five training datasets represents a distinct set of features, which are used to learn five different classification models. (c) Combination level—the prediction results generated from the five models are combined to obtain the final drug sensitivity predictions. CNV, copy-number variation; RPPA, reverse phase protein array.