| With the aging of the population and the increasing incidence of dementia,Alzheimer’s disease(AD)has become a serious public health problem.As a slowly progressing neurodegenerative disease,the pathological changes of AD have occurred ten or even twenty years before the appearance of clinical symptoms,leading to the lack of specificity and sensitivity in the clinical diagnosis of AD.The development of brain imaging methods and AD biomarkers provide powerful technical means for the detection of AD-related pathologies.The study of twins has proved that AD has high heritability.Therefore,using AD brain imaging biomarkers as the internal phenotype to study their genetic associations with genomic data,and then analyzing genetic variations and their influence on brain function,has become an effective way to reveal the pathological mechanism of AD and promote early diagnosis and intervention of AD.Based on the "non-hypothesis driven" advantages of genome-wide association study(GWAS),we investigate the network functional modules and trans-omic cascade effects at the locus level,gene level,module level,and multi-omic level,aiming to discover multi-level genetic variations that are significantly associated with AD quantitative imaging phenotypes,so as to provide new biological insights for revealing the polygenic basis of disease susceptibility.The main research contents and contribution of this paper are as follows:(1)In this paper,a genetic association analysis strategy based on quantitative imaging phenotypes is designed to identify the genetic associations between genotypes and AD quantitative imaging phenotypes at the locus level.Based on the quality control of AD quantitative imaging phenotype and genotype data,we construct the genetic correlation model between brain imaging and genome based on linear regression model,and introduce multiple hypothesis test,population structure analysis and sample distribution test to statistically evaluate the GWAS results.The results show that the quality control process is reasonable and the genetic association framework given is effective.(2)Based on the biological characteristics that protein complexes and the internal proteins in a functional module tend to be closely related,we propose a local network functional module mining strategy based on graph density for the AV-45 PET imaging phenotype that characterizes the AD amyloid pathology,so as to identify dense network functional modules that are significantly enriched by genetic variations with moderate effects.This strategy integrates the significance of genetic variations and their network neighborhood topology features to identify modules that are significantly associated with the AV-45 PET imaging phenotype,and introduces topology bias correction to reduce potential deviations of the network topology.A random significance correction process based on Monte Carlo simulation is proposed to eliminate the false positive results introduced by linkage disequilibrium.Replication experiments with independent samples increase the reliability of the results.The importance of risk genes is assessed by network topology analysis and module similarity measurement.The biological significance of candidate modules is evaluated through functional enrichment analysis and biological experiment literature mining.The results demonstrate the effectiveness of the GWAS-based dense network functional module mining strategy in discovering the potential disease phenotype-associated signals with individual moderate effects.(3)Not all protein complexes have dense graph structures.Some genetic signals of complex diseases have fewer known interactions because they are not widely studied.The mining of non-dense network modules will help to discover these peripheral nodes that affect complex diseases and to generate new genetic hypotheses.In addition,the signals and noises introduced by different biological network backgrounds are different,resulting in the lack of consistency in the identified network functional modules.To this end,a non-dense concencus hierarchical network functional module mining strategy is proposed for the hippocampal MRI imaging phenotype that characterizes the AD brain atrophy pathology.After the dynamic construction of the similarity matrix,this strategy gives a reasonable division of network hierarchy based on the maximization of deviation.A consensus hierarchical network module mining strategy that weighs the sensitivity and specificity of different biological networks is proposed,so as to reduce the false positives caused by potential network topology bias.The correlation between candidate modules and AD pathology is verified from multiple perspectives,such as topological feature calculation,GO functional annotation,enrichment analysis,and biological experimental literature mining.The results show that the GWAS-based non-dense concencus hierarchical network functional module mining strategy is effective for sparse connections,and verify the joint effects of non-dense interaction of genes with individual moderate effects on the disease phenotype.(4)Due to the lack of comprehensiveness and heterogeneity in the systematic characterization of complex trait biological models,a trans-omic enrichment method for disease biomarkers based on heterogeneous multi-omic biologic network is proposed in this paper.This method constructs a heterogeneous multi-omic network model that characterizes the cascading effects from genetic variants to genes to biochemical reactions and finally to downstream metabolites.By identifying disease-related trans-omic cascade effects among different omic layers of genome,proteome,and metabolome,this method performs functional annotation of disease risk variants and prediction of downstream disease pathology-related metabolites.The results show that the construction of heterogeneous multi-omic biological network model plays a potential role in the molecular pathway of disease disorders,and verify the effectiveness of heterogeneous multi-omic network features in biomarker enrichment analysis. |
PDF Full Text Request