Heritability estimation of reliable connectome features

Abstract

Brain imaging genetics is an emerging research field aimed at studying the underlying genetic architecture of brain structure and function by utilizing different imaging modalities. However, not all the changes in the brain are a direct result of the genetic effect. Furthermore, the imaging phenotypes are promising for genetic analyses are usually unknown. In this thesis, we focus on identifying highly heritable measures of structural brain networks derived from Diffusion Weighted Magnetic Resonance imaging data. Using data for twins that is made available by the Human Connectome Project (HCP), the reliability of edge-level measures, namely fractional anisotropy, fiber length, and fiber number in the structural connectome, as well as seven network-level measures, specifically assortativity coefficient, local efficiency, modularity, transitivity, cluster coefficient, global efficiency, and characteristic path length, were evaluated using intraclass correlation coefficients. In addition, estimates of the heritability of the reliable measures were also obtained. It was observed that across all 64,620 network edges between 360 brain regions in the Glasser parcellation, approximately 5% were significantly high heritability based on fractional anisotropy, fiber length, or fiber number. Moreover, all tested network level measures, that capture network integrity, segregation, or resilience, were found to be highly heritable, having a variance ranging from 59% to 77% that is attributable to an additive genetic effect.