Figure 1.

Overview of the three applied model building strategies. (a) Use of a single data set; (b) manual integration of data over time; (c) a genome-wide integration approach. The data sets are represented as matrices with rows corresponding to patients and columns corresponding to genes, proteins, or CNVs. In step A, LS-SVM models are built on each data set separately. A two-dimensional grid is used for the optimization of the regularization parameter and the number of features. For step B, data sets over time are combined. By using the changes in expression or abundance as features, a two-dimensional grid is suficient. In step C, an intermediate integration method is used for the integration of all available data sets. A k-dimensional grid is required for optimizing the regularization parameter and the number of features selected from the (k - 1) integrated data sets. FS, feature selection; M_i, model for parameter combination i; NF, number of features; T, time point.