PO-460 Gene expression-based probabilistic graphical models identify three independent biological layers in colorrectal cancer

IntroductionColorectal cancer (CRC) is the fourth cause of cancer death in the world. The Colorectal Cancer Subtyping Consortium (CRCSC) defined four consensus molecular subtypes (CMS): immune, canonical, metabolic and mesenchymal. This study bases on the hypothesis that there are two independent biological layers, as our group has previously seen in other malignancies, such as Triple negative breast cancer, melanoma or bladder cancer. So, the aim of this studio is to put the colorrectal cancer subtypes in context with that hypothesis.Material and methodsColorectal cancer tumour datasets were obtained from Gene Expression Omnibus (GEO). A merged database was constructed, and limma R package was used to remove batch effect. A probabilistic graphical model, followed by successive sparse k-means and consensus cluster analyses were used to classify CRC tumour samples from the merged dataset.Results and discussionsThrough this approach, three independent classifications, immune, molecular and adhesion-related, were established. Molecular classification showed different groups of tumours that present differential expression of diverse colorectal biomarkers. The immune classification further divided patients into high immune activity and low immune activity subgroups. And the adhesion-related assignment divides patients in two subgroups with prognostic value. We showed that immune and adhesion-related layers, when studied together, add its prognostic value defining low risk and high risk subgroups. We finally compare our classifications with the consensus molecular classification.ConclusionMolecular information obtained using a novel analytical approach to classify colorectal cancer tumour samples allowed us to establish new classifications based on different biological layers. We have demonstrated that our classifications based on the immune and the adhesion-related layers have prognostic value. Ultimately, to deepen the knowledge of CRC makes it possible to propose new therapeutic strategies and could also be used for diagnostic and prognostic purposes.