An international consortium of colorectal cancer researchers undertakes a large-scale data sharing project to achieve a consensus molecular classification of colorectal cancers. 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 (CMSs) with distinguishing features: CMS1 (microsatellite instability immune, 14%), hypermutated, microsatellite unstable and strong immune activation; CMS2 (canonical, 37%), epithelial, marked WNT and MYC signaling activation; CMS3 (metabolic, 13%), epithelial and 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 intratumoral 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.

Reactivating the fetal globin gene: Mutation of adult-type globin genes causes sickle cell disease and thalassemia. Although treating these hemoglobinopathies with gene therapy is possible, there is a pressing need for pharmacologic approaches to treat general patient populations. One promising approach is to reactivate repressed expression of fetal-type hemoglobin (HbF) in adult erythroid cells. Masuda et al. reveal a molecular mechanism governing HbF repression as mediated by the LRF/ZBTB7A transcription factor. The study may encourage the development of new HbF reactivation therapies for hemoglobinopathies. Science , this issue p. 285

The fibroblast growth factor receptor 2 (FGFR2) locus has been consistently identified as a breast cancer risk locus in independent genome-wide association studies. However, the molecular mechanisms underlying FGFR2-mediated risk are still unknown. Using model systems we show that FGFR2-regulated genes are preferentially linked to breast cancer risk loci in expression quantitative trait loci analysis, supporting the concept that risk genes cluster in pathways. Using a network derived from 2,000 transcriptional profiles we identify SPDEF, ERα, FOXA1, GATA3 and PTTG1 as master regulators of fibroblast growth factor receptor 2 signalling, and show that ERα occupancy responds to fibroblast growth factor receptor 2 signalling. Our results indicate that ERα, FOXA1 and GATA3 contribute to the regulation of breast cancer susceptibility genes, which is consistent with the effects of anti-oestrogen treatment in breast cancer prevention, and suggest that fibroblast growth factor receptor 2 signalling has an important role in mediating breast cancer risk.

Molecular subtyping of cancer is a critical step towards more individualized therapy and provides important biological insights into cancer heterogeneity. Although gene expression signature-based classification has been widely demonstrated to be an effective approach in the last decade, the widespread implementation has long been limited by platform differences, batch effects, and the difficulty to classify individual patient samples. Here, we describe a novel supervised cancer classification framework, deep cancer subtype classification (DeepCC), based on deep learning of functional spectra quantifying activities of biological pathways. In two case studies about colorectal and breast cancer classification, DeepCC classifiers and DeepCC single sample predictors both achieved overall higher sensitivity, specificity, and accuracy compared with other widely used classification methods such as random forests (RF), support vector machine (SVM), gradient boosting machine (GBM), and multinomial logistic regression algorithms. Simulation analysis based on random subsampling of genes demonstrated the robustness of DeepCC to missing data. Moreover, deep features learned by DeepCC captured biological characteristics associated with distinct molecular subtypes, enabling more compact within-subtype distribution and between-subtype separation of patient samples, and therefore greatly reduce the number of unclassifiable samples previously. In summary, DeepCC provides a novel cancer classification framework that is platform independent, robust to missing data, and can be used for single sample prediction facilitating clinical implementation of cancer molecular subtyping.

NUT midline carcinoma (NMC), a subtype of squamous cell cancer, is one of the most aggressive human solid malignancies known. NMC is driven by the creation of a translocation oncoprotein, BRD4-NUT, which blocks differentiation and drives growth of NMC cells. BRD4-NUT forms distinctive nuclear foci in patient tumors, which we found correlate with ∼100 unprecedented, hyperacetylated expanses of chromatin that reach up to 2 Mb in size. These “megadomains” appear to be the result of aberrant, feed-forward loops of acetylation and binding of acetylated histones that drive transcription of underlying DNA in NMC patient cells and naïve cells induced to express BRD4-NUT. Megadomain locations are typically cell lineage-specific; however, the cMYC and TP63 regions are targeted in all NMCs tested and play functional roles in tumor growth. Megadomains appear to originate from select pre-existing enhancers that progressively broaden but are ultimately delimited by topologically associating domain (TAD) boundaries. Therefore, our findings establish a basis for understanding the powerful role played by large-scale chromatin organization in normal and aberrant lineage-specific gene transcription.