Identification Of Related Gene Sets Based On Multigene Expression Pattern Analysis

As the rapid development of high-throughput sequencing,researchers can perform sequencing for the genomes of human individuals from different groups to study the differences between groups.Currently,sets of single nucleotide variations(SNVs)and copy number variations(CNVs)related to neuropsychiatric disorders have been identified from genetic studies,in particular genome-wide association studies(GWAS).However,the identification of actual causal genes is still difficult,because the identified SNVs are often located in intergenic regions and CNVs often involve more than one variants or genes.Measuring and analyzing gene expression in the primary affected organ by neuropsychiatric disorders directly,namely brains,promotes to reveal the influence of regulatory variants on the gene itself and the expression of other genes,allowing for novel insights in understanding neuropsychiatric disorders.Same to this,one can also study normal brain activity,for instance,aging,by analyzing the gene expression in brains to access the potential relationship between age and gene expression.RNA-seq is a kind of transcriptome sequencing which uses deep sequencing technology,allowing us to accurately and quantitatively measure the gene transcription product,mRNA,and then determine the expression level of genes.The increasing of tissue samples of human brains makes it possible to perform RNA sequencing for these tissue samples and then analyze the level of gene expression for studying further the normal brain aging and neuropsychiatric diseases.On studying the differences between different groups of individuals(such as diseased and control groups,different age groups)based on gene expression information,more study designs mainly focus on the difference in expression means between groups,namely differential expression(DE)analysis.The main problem is,based on its nature,group difference-based method will blur or ignore the heterogeneity within groups and then may ignore genes with variable expression level within groups.Particularly,for the complex disorders with phenotypic and genetic heterogeneity,DE analysis may overlook the expression variability among affected individuals or samples by diseases caused by genetic heterogeneity.Differential variability(DV)analysis was developed to capture the difference in expression variability between different groups,which can be used to detect genes with significantly different variance of expression level between groups.In addition,as genes are usually correlated,we also need to identify gene sets associated with the studied phenomenon or disease,whose multigene expression patterns are significantly different between groups.At present,to study normal aging and neuropsychiatric disorders,the methods using multivariant analysis to analyze gene expression data are still not enough.The main content of this thesis is to identify sets of genes associated with aging or neuropsychiatric disorders based on the analysis of multigene expression pattern.1.Aging shapes gene expression pattern in human brainsTo study the potential relationship between age and gene expression in human brains,we analyzed the gene expression in tissue samples from 13 brain regions included in Genotype-Tissue Expression(GTEx).We identified the protein-coding genes whose expression are correlated with age for each brain region.Using dispersion-specific analysis,we identified individual genes and gene sets whose expression levels are differentially dispersed by aging.Our analysis results show age associated gene expression is specific to brain region,and related to changes of gene expression mean and dispersion.2.Aberrant gene expression analysis for a single diseaseTo identify gene sets related to diseases with substantial phenotypic and genetic heterogeneity,such as autism(AUT),we developed aberrant gene expression analysis.We analyzed gene expression data of brain tissues from AUT and controls,and identified 54 gene sets and 76 co-expression modules whose expression are dispersed significantly in AUT samples.With a gene expression dataset from whole blood,we identified three aberrantly expressed gene sets which can be used for AUT diagnosis,achieving>70%classification accuracy.In addition,a web service and a stand-alone software were implemented for identifying gene sets associated with a specific disease.3.Identification of shared gene sets across multiple diseasesTo understand the downstream impact of genetic overlap on gene expression between AUT,schizophrenia(SCZ)and bipolar disorder(BPD),we applied two kinds of multivariant analysis methods including gene co-expression network analysis and aberrant gene expression analysis.We analyzed the gene expression of AUT,SCZ,BPD and controls to identify shared gene sets between these three disorders for studying the extent of similarity between gene expression of AUT,SCZ and BPD.Based on the analysis of gene expression pattern,we identified individual genes or gene sets associated with aging or neuropsychiatric disorders.In the study of aging shaping the gene expression pattern in human brains,we not only considered the gene expression mean but also dispersion,providing a foundation for more complicated modeling of gene expression in studying neurodegenerative diseases related to age.In the gene expression analysis for neuropsychiatric disorders,we not only used the traditional gene co-expression network analysis,but also developed aberrant gene expression analysis method considering the expression heterogeneity within group of individuals affected by a disease.Our result provides new insights for studying the genetic and molecular mechanisms underlying the gene expression dysregulated in diseased individuals.