Component-based methods for data reduction and integration: application to genomics and neuroimaging studies.

点击翻译按钮获取中文摘要

It is now well accepted that the biggest challenges and upcoming breakthroughs in the genomic field will be derived from handling big data. Not only is big data generated faster than our own capacity to fully understand, interpret and evaluate the results produced, it is also collected from an ever-increasing number of sources. As a result, statistical challenges with integrating this information into a unified model of genome function needs to be addressed. With a particular focus on genomic studies, the proposed research endeavours to develop a set of new statistical tools designed to investigate association between two or more sets of possibly high dimensional correlated variables measured on the same subjects. In particular, the general objective of the proposal is to tackle both the issue of data reduction and association analysis simultaneously, by developing statistical models achieving an optimal data reduction towards the final goal of detecting associations between datasets. Graphical model tools, such as the ones discussed in this project, need to be carefully evaluated in the context of genomic studies. The potential of such models to uncover new complex network patterns is tremendous and recent development in this fields has led to an to increase in their computing efficiency. This proposal has also been developed in the context of the creation of a new institute the Ludmer Centre for Neuroinformatics and Mental Health at McGill. Researchers at the new Ludmer Centre engage in a special multidisciplinary research platform, with the goal of understanding how different genetic, epigenetic and environmental factors influence brain development in children. The approach combines disciplines including neuroscience, computational biology, mathematics, genetics, epigenetics, bioinformatics, epidemiology and computer science. In this context, the proposed project is very relevant and will therefore enhance direct knowledge transfer. In particular, I expect that discoveries will open new pathways for diagnosis, prevention and treatment. **