Constructing genotype and phenotype network helps reveal disease heritability and phenome-wide association studies

Abstract

Analyses of a bipartite Genotype and Phenotype Network (GPN), linking the genetic variants and phenotypes based on statistical associations, provide an integrative approach to elucidate the complexities of genetic relationships across diseases and identify pleiotropic loci. In this study, we first assess contributions to constructing a well-defined GPN with a clear representation of genetic associations by comparing the network properties with a random network, including connectivity, centrality, and community structure. Next, we construct network topology annotations of genetic variants that quantify the possibility of pleiotropy and apply stratified linkage disequilibrium (LD) score regression to 12 highly genetically correlated phenotypes to identify enriched annotations. The constructed network topology annotations are informative for disease heritability after conditioning on a broad set of functional annotations from the baseline-LD model. Finally, we extend our discussion to include an application of bipartite GPN in phenome-wide association studies (PheWAS). The community detection method can be used to obtain a priori grouping of phenotypes detected from GPN based on the shared genetic architecture, then jointly test the association between multiple phenotypes in each network module and one genetic variant to discover the cross-phenotype associations and pleiotropy. Significance thresholds for PheWAS are adjusted for multiple testing by applying the false discovery rate (FDR) control approach. Extensive simulation studies and analyses of 633 electronic health record (EHR)-derived phenotypes in the UK Biobank GWAS summary dataset reveal that most multiple phenotype association tests based on GPN can well-control FDR and identify more significant genetic variants compared with the tests based on UK Biobank categories.