Colorectal cancer (CRC) is increasingly being recognized as a heterogeneous disease with distinct molecular subtypes. These subtypes have different biological processes at the basis of their disease and consequently their prognosis and responses to therapy are also different. We have previously developed molecular diagnostic assays using a single platform on routine FFPE tumour biopsies. These assays identify gene expression profiles with distinct prognosis and drug response phenotypes (CMS4/c-type, BRAF mutant-like, and MSI-like). Our overall objective is to develop targeted therapies more effective than the current therapies that do not take advantage of molecular classification of the disease to select patients for therapy. We therefore propose to perform 3 two-stage single arm multi-centre open-label phase II studies based on solid preclinical evidence and a sound scientific rationale for these subgroups of CRC patients: 1) combination of chemotherapy and a TGF-βR inhibitor (LY2157299) in patients presenting a C-type signature; 2) vinorelbine in patients with a BRAFm-like signature; and 3) an immunotherapeutic anti-PD-L1 drug (MPDL3280A) in combination with bevacizumab in patients with a MSI-like signature. The primary objectives of these studies are to determine the clinical efficacy (progression-free survival as primary endpoint), safety and tolerability of the experimental treatments in these molecularly selected populations. Mutation analysis at the beginning of treatment and monitoring by liquid biopsies might reveal further biomarkers that predict response in retrospective analysis. The project outcomes may have a significant impact in CRC patients with poor-risk prognosis worldwide as 40-50% of them present gene expression profiles matching one of the 3 approaches. Around 40,000 European CRC patients may potentially benefit from the results. Also, these may be translated to other cancer types with equivalent gene expression patterns/deregulated signalling pathways.