Premium
Functional molecular network models for the genetic risk factors of Alzheimer’s disease
Author(s) -
Ming Chen,
Marcora Edoardo,
Wang Minghui,
Renton Alan E.,
Wang Erming,
Goate Alison,
Zhang Bin
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.046556
Subject(s) - genome wide association study , biology , genetic association , genetics , disease , computational biology , gene , single nucleotide polymorphism , medicine , genotype
Background Genome‐wide association studies (GWAS) have pinpointed a limited number of risk loci in Alzheimer’s disease (AD). However, the biological functions of these AD risk loci remain mostly elusive. Network analysis of functional genomic data provides an excellent opportunity to understand AD systematically. In this study, we leverage a cohort of large‐scale transcriptomic data in AD to develop functional network models for AD risk genes. Method We integrated all sumstats of AD risk loci from four previous AD GWAS studies and generated putative AD‐risk genes(PADRG) using MAGMA. We determined the +/‐ 1 Mb cis‐regulome interval for the promoter of each protein‐coding gene, and applied MAGMA in top SNP mode to all AD GWAS summary statistics using cis‐regulome intervals as the unit of aggregation, then retained genes whose interval achieved both p < 5e‐8 in any GWAS and p < 1 in all four GWAS. Based on the Mount Sinai Brain Bank AD cohort comprised of multi‐Omics data from multiple brain regions, we performed the multiscale embedded gene coexpression network (MEGEN) analysis to delineate gene coexpression relationships. Subnetworks centering around each PADRG were then extracted and evaluated for relevance with AD‐related clinical and pathological traits. Result The 751 PADRGs are enriched in the immune response‐related modules in the MEGENs from four brain regions, consistent with the known prominent role of immune response in AD. The subnetwork analysis further identifies the functional contexts of PADRGs. All the top thirty‐nine PADRG‐centered subnetworks are from the parahippocampal gyrus region and have a variety of biological functions such as synaptic signaling, protein‐complex subunit organization, and RAS protein signal transduction. The top two subnetworks are centered around FAM81A and RCAN2 and are involved in γ‐aminobutyric acid (GABA) signaling. Three well‐known AD risk genes CLU , MAPKAPK2 and STX3 , are predicted to function in the cellular response to organic substance, inflammatory response, synaptic signaling pathways, respectively. Conclusion Our multiscale network analysis systematically identifies the functional contexts of the PADRGs. The subnetworks around the PADRGs and the comprehensive characterization of those subnetworks lay down a solid foundation for not only prioritizing AD‐associated genes but also investigating their biological functions.