Background, rationale: Gliomas, the most common primary brain tumors, are divided into low-grade and high-grade gliomas. Wild-type IDH-1 gliomas share aggressive biological behaviour and poor prognosis. IDH-1 mutated gliomas are characterized by a less aggressive biological behaviour, a clinical prognosis less influenced by tumor grade, and include diffuse low grade and intermediate grade gliomas, recently defined as Lower Grade Gliomas (LGGs). This project focuses on LGGs and how to integrate imaging and molecular heterogeneity into a tool to be used in the routine clinical practice for easily predicting LGGs progression. Predicting the prognosis using such high dimensional and heterogeneous data required specific approaches.
Hypothesis: Tumor evolution is of primary importance in IDH1-mutated LGG regarding clinical outcome. Global genetic profiling of the primary tumors is not sufficient. A detailed multi-layer analysis needs to be undertaken. We postulate that an integrative analysis of imaging, transcriptional, proteomic data coupled with molecular patient data and immunohistology will provide a better understanding of the clinical evolution of IDH1-mutated LGG. It is to emphasize that our approach is a multilevel approach, where we aim at obtaining insights about the heterogeneity profiles of the tumor. The aim is to integrate the data by computational modelling for prediction of clinical evolution of the disease.
Aims: The aim is to elaborate a predictive model for LGG progression by using a novel approach coupling mathematical modelling and statistical learning adapted to high dimensional data. The models will be further refined by combining standard molecular analysis as well as expression analysis, proteomics and infrared imaging.
Methods: The general approach of the project will be divided in several steps: i) Construction of the predictive model v1.0 of clinical progression based on standard clinical, molecular and imaging characteristics in a cohort of 150-200 patients followed over up to 5 years ii) Deep evaluation of the misclassified patients based on further imaging, transcriptomics and proteomics iii) Supervised analysis of each high-dimensional datasets based on appropriate and new mathematical and statistical models iv) Construction of the predictive model v2.0 improving the model 1.0 with the transcriptomics and proteomics analyses v) Validation of the model v2.0 using a test sample.
Expected results and potential impact: A predictive model for patient stratification and prediction at onset of diagnosis is the ultimate deliverable that should improve the clinical management of patients. Additional results should be provided by the multilevel approach leading to a better understanding of the disease.