Neuroimaging enables the characterization of subtle alterations of brain structure and activity. Those characteristics can be considered as quantitative traits or endophenotypes that can be analyzed together with genomics. A well accepted hypothesis is that the modification of the brain structure/activity is modulated by genomic biomarkers. Therefore, the close study of correlations between genomics and neuroimaging datasets enables one to identify genes implicated in the alterations of a given neuroimaging structure and to understand the underlying biological mechanisms at work for a given disease. From this perspective, we propose to mine a cohort of patients with Spinocerebellar Ataxia disorder (SCA – a neurodegenerative disease) on which multimodal genomics datasets (expression and metabolomics data), multimodal neuroimaging (volumetry, magnetic resonance spectroscopy) and clinical information have been acquired. All participants (67 patients / 35 controls) came for a baseline assessment and then 2 years later for a follow-up visit.
The joint analysis of these complex and heterogeneous sources of information promises to significantly improve our understanding of the molecular mechanisms of SCA. The goal of this integrative analysis is primarily to identify a set of genomic biomarkers co-varying with a set of neuroimaging variables (e.g. cerebellar and brainstem tissue loss) which are associated with disease severity/progression. The main bottlenecks of such an integrative analysis are the high complexity and heterogeneity of the data that stems from: (i) various sources, (ii) the number of variants of SCA to be considered (iii) the high number of measurements in both genomics and neuroimaging data which involves the computation of million(s) of associations. Consequently, a successful investigation of such a structured dataset requires to develop/use statistical methods that fit both the peculiar structure of the data as well as their heterogeneous nature.
All the analyses might be conducted based on the recently published methods [Tenenhaus et al 2015, Lofstedt et al 2015, Tenenhaus & Tenenhaus 2014a, Tenenhaus et al 2014b, Tenenhaus & Tenenhaus 2011]. The methodological tools developed during this project will lead to prototype software (e.g. R packages).
All the datasets are already centralized at the bioinformatics and biostatistics platform of the ICM.