Colorectal cancer (CRC) is a frequently lethal disease with heterogeneous outcomes and drug responses. To resolve inconsistencies among the reported gene expression–based CRC classifications and facilitate clinical translation, we formed an international consortium dedicated to large-scale data sharing and analytics across expert groups. We show marked interconnectivity between six independent classification systems coalescing into four consensus molecular subtypes (CMS) with distinguishing features: CMS1 (MSI Immune, 14%), hypermutated, microsatellite unstable, strong immune activation; CMS2 (Canonical, 37%), epithelial, chromosomally unstable, marked WNT and MYC signaling activation; CMS3 (Metabolic, 13%), epithelial, evident metabolic dysregulation; and CMS4 (Mesenchymal, 23%), prominent transforming growth factor β activation, stromal invasion, and angiogenesis. Samples with mixed features (13%) possibly represent a transition phenotype or intra-tumoral heterogeneity. We consider the CMS groups the most robust classification system currently available for CRC – with clear biological interpretability – and the basis for future clinical stratification and subtype–based targeted interventions.

Next-generation sequencing (NGS) allows sequencing of a high number of nucleotides in a short time frame at an affordable cost. While this technology has been widely implemented, there are no recommendations from scientific societies about its use in oncology practice. The European Society for Medical Oncology (ESMO) is proposing three levels of recommendations for the use of NGS. Based on the current evidence, ESMO recommends routine use of NGS on tumour samples in advanced non-squamous non-small-cell lung cancer (NSCLC), prostate cancers, ovarian cancers and cholangiocarcinoma. In these tumours, large multigene panels could be used if they add acceptable extra cost compared with small panels. In colon cancers, NGS could be an alternative to PCR. In addition, based on the KN158 trial and considering that patients with endometrial and small-cell lung cancers should have broad access to anti-programmed cell death 1 (anti-PD1) antibodies, it is recommended to test tumour mutational burden (TMB) in cervical cancers, well-and moderately-differentiated neuroendocrine tumours, salivary cancers, thyroid cancers and vulvar cancers, as TMB-high predicted response to pembrolizumab in these cancers. Outside the indications of multigene panels, and considering that the use of large panels of genes could lead to few clinically meaningful responders, ESMO acknowledges that a patient and a doctor could decide together to order a large panel of genes, pending no extra cost for the public health care system and if the patient is informed about the low likelihood of benefit. ESMO recommends that the use of off-label drugs matched to genomics is done only if an access programme and a procedure of decision has been developed at the national or regional level. Finally, ESMO recommends that clinical research centres develop multigene sequencing as a tool to screen patients eligible for clinical trials and to accelerate drug development, and prospectively capture the data that could further inform how to optimise the use of this technology.

Critical driver genomic events in colorectal cancer have been shown to affect the response to targeted agents that were initially developed under the 'one gene, one drug' paradigm of precision medicine. Our current knowledge of the complexity of the cancer genome, clonal evolution patterns under treatment pressure and pharmacodynamic effects of target inhibition support the transition from a one gene, one drug approach to a 'multi-gene, multi-drug' model when making therapeutic decisions. Better characterization of the transcriptomic subtypes of colorectal cancer, encompassing tumour, stromal and immune components, has revealed convergent pathway dependencies that mandate a 'multi-molecular' perspective for the development of therapies to treat this disease.

The phosphatidylinositol 3-kinase (PI3K) pathway has an important role in cell metabolism, growth, migration, survival and angiogenesis. Drug development aimed at targetable genetic aberrations in the PI3K/AKT/mTOR pathway has been fomented by observations that alterations in this pathway induce tumour formation and that inappropriate PI3K signalling is a frequent occurrence in human cancer. Many of the agents developed have been evaluated in early stage clinical trials. This Review focuses on early clinical and translational data related to inhibitors of the PI3K/AKT/mTOR pathway, as these data will likely guide the further clinical development of such agents. We review data from those trials, delineating the safety profile of the agents--whether observed sequelae could be mechanism-based or off-target effects--and drug efficacy. We describe predictive biomarkers explored in clinical trials and preclinical mechanisms of resistance. We also discuss key unresolved translational questions related to the clinical development of inhibitors of the PI3K/AKT/mTOR pathway and propose designs for biomarker-driven trials to address those issues.