Abstract
Alternative splicing is one of the major mechanisms leading to a diversity in the proteome. It has become very clear that this mechanism is playing a role in many genetic diseases including cancer. During oncogenesis, the cellular content of RNA isoforms is highly altered and this phenomenon seems to be context specific. Even in the same tissue, the pool of transcripts can display specific rearrangements corresponding to different subtypes of the disease.As we know now, the majority of deaths from solid tumors are caused by metastases. This metastatic cascade might involve the Epithelial Mesenchymal Transition (EMT) which is a complex biological trans-differentiation process that allows epithelial cells to transiently obtain mesenchymal features. During this process, an alternative splicing program is differentially regulated, and increasing number of studies have started to suggest that a simple isoform switching is sufficient to start an EMT. Stopping the spreading of cancer cells in the human body represent an important challenge in the fight against cancer. In this context, we believe that exploring alternative splicing could add a thinner layer or regulation to classify patients more precisely, help to discover new potential targets for therapy and therefore, improve patient care. As we are in the era of precision medicine, we use a rational approach, focusing a specific subtype to avoid bias due to heterogeneity of breast cancer. We focus on basal-like breast cancer which is one of the most aggressive and deadly among all breast cancers.During this work, we took advantage of datasets from a large-scale initiative (The Cancer Genome Atlas) which provides researchers with cancer genomics and associated clinical follow-up. We analyzed 188 patients with basal-like breast cancer for gene expression and alternative splicing. Based on breast cancer cell lines RNA-sequencing, using a custom machine learning approach based on Random Forest, we succeeded to distinguish two groups of patients with distinct prognosis. Using several public EMT-induced RNA sequencing projects, we confirmed these alternative splicing events were linked to EMT.As a side project, we also got involved in the development of methods of classification using k-mers. We first were involved in a project that test the ability of k-mer to classify breast cancer subtype. Secondly, we were focused in the discovery of biological knowledge that k-mers are bringing in the breast cancer stratification.Finally, our results show that alternative splicing or k-mers can be the source of new valuable information to help in the thinner definition of oncogenic subtypes or identification of biological processes in cancer. In a breast cancer subtype that does not benefit from targeted therapy, we demonstrate that alternative splicing relative to an EMT could be used as potential biomarkers to isolate patients where the tumor progress faster. This work could help to develop new treatments for precision oncology.