B0219
Title: Network enrichment significance testing in brain-behavior association studies
Authors: Sarah Weinstein - Temple University (United States) [presenting]
Abstract: In neuroimaging studies, functional parcellations of the brain are widely used as a framework for interpreting brain-behavior associations. Despite the widespread use of these parcellations in neuroimaging analysis, there remains no common framework for testing network "enrichment"; that is, whether or not observed brain-behaviour associations are especially strong within a network of interest. A framework for network enrichment significance testing is described. The method is a generalization of gene set enrichment analysis (GSEA), widely used in the genomics literature to test enrichment between sets of genes and phenotypes. The extension of GSEA to the neuroimaging context takes a quantitative brain map (where each value measures an association of interest at each location) and a subset of locations in the brain (for instance, a set of vertices on the cortical surface that form the default mode network). The method then returns an enrichment score, which quantifies the degree to which associations within versus outside that subregion are strong. Permutation is then used for inference. The method is then applied in a large-scale study of neurodevelopment, illustrating its flexibility and power through data-driven simulations and real data analysis, involving associations between measures of brain structure and neurodevelopmental phenotypes.
