Fig. 1

Overview of the multi-task learning method. RNA-Seq data from drug response experiments in different cell lines or cancer types (top) is used to select pathways linking source proteins to DE genes in general protein-protein and protein-DNA interaction networks (second row). Reconstructed networks are constrained by encouraging pathways that are shared across different cancer types leading to a general network (third row) that captures the common pathways activated during the response. In addition to the general network, cell type specific networks are also identified (bottom) and these can help identify tissue specific proteins and explain differences in response of certain cancer types when treated with the same drug