Mots-Clés

Omics data, machine-learning, colorectal cancer, classification

Description

Colorectal cancer (CRC) is the third most common cancer worldwide and metastatic colorectal cancer (mCRC) is one of the leading causes for cancer-related death. Indeed, 25-50% of patients with CRC will develop metastatic disease, which ultimately results in death for more than two thirds of the patients. The hepatic resection for metastatic tumors from CRC remains the only curative option, and systemic or intra-arterial hepatic chemotherapy constitutes an alternative in some cases for patients with unresectable disease. Thereby, refining patient selection is essential to improving treatment chose and outcome. Thus, a multidisciplinary approach for patient stratification will help to determine the optimal therapeutic options for each patient. In 2015, an international consortium of experts identified, from more than 4x103 CRC sample patients, four transcriptomic consensus molecular subtypes (CMS) characterized by molecular, biological and clinical factors. To date, this classification has demonstrated prognostic value in several situations but is not yet routinely used in clinical practice. Indeed, different CMS subtypes can be represented in the same tumor and all the CMS subtypes do not present a clear boundary between groups.

We propose that combining this classification to proteomic elements could make it possible to generate an extensive identity card of a tumor, and establish a multimodal classification to characterize the proteomic profile of each CMS that reinforce the predictive value. We will be able to classify patients into separate groups according to their genetic/epigenetic and proteomic features which could help with choosing the best therapeutic approach and the development of a precision medicine strategy. The combination of CMS and proteomic profiling for personalized target identification is therefore a promising avenue to address the need for new effective therapeutic tools for metastatic CRC, particularly the unrespectable mCRC.

The integration of proteomic and transcriptomic data obtained from a selection of 20 and 60 patients, respectively, with known phenotypes will be the focus of the bioinformatics internship. The data are originating from normal colon tissues, primary and metastatic tumors of the same patients that had surgery as initial therapy. The aim will be to compare different statistics and machine learning methods for integrating these data and identifying features for classifying patients.

The successful candidate will characterize these signatures related to the CMS types and will evaluate the biological relevance of the list of genes/proteins.

The internship will take place at the LaBRI, within an interdisciplinary environment. A knowledge of basic bioinformatic skills (python, R, Bash…etc) is required. Good communication and presentation skills are necessary. The candidate will work in close collaboration with the LaBRI et BRIC.