Genetics effect of structural brain network topology based on network characterization

Abstract

Imaging genetics has become a popular research area among other objectives, to quantify genetic factors (heritability) and non–genetic factors contributing to variations in brain network both in neurodevelopment and psychiatric disorders. Prior to this study, few studies utilizing Diffusion Tensor Imaging (DTI) used only binary network (BNs) for heritability analysis of structural brain network. However, the procedures for constructing binary network are not without limitations. It was hypothesised that if network parameters are truly heritable, heritability estimates of these parameters from the two most basic strategies of constructing network in neuroimaging studies should give comparable results. Therefore, this current study sought to determine whether the estimated heritability values utilizing binary networks are reliable, consistent or robust to methodological choices by replicating the results with weighted network (WNs) using the twin study design. Paired-difference T-Test results for all network metrics considered under this study produced significant differences (p < 0.05) between WNs and BNs estimates. This confirms that BNs and WNs represent different classes of network. Heritability estimates of widely used network metrics in the literature: normalized clustering coefficient (γ), global efficiency (λ), characteristic path length () and small–worldness (σ) of structural brain network were estimated. All the aforementioned network metrics were moderately heritable (40% - 60%) with slightly differences between WNs and BNs. In our quest to search for genes influencing the human brain network, this study demonstrates the reliability or consistency of estimated heritability values of structural brain network. Regardless of the network construction strategy employed, heritability estimates of all the metrics were moderately heritable. These findings imply that network parameters represent promising or reliable endophenotypes for future Genome Wide Association Studies or Quantitative Trait Loci (QTL) analyses which intend to search for susceptibility genes influencing the human brain. Moreover, this study provides useful information for future meta-analysis studies because the validity of meta-analysis by combining BNs and WNs is not in doubt due to the findings in this study.