The past few decades have seen dramatic increase in the prevalence of metabolic diseases (MetDs) including obesity, type 2 diabetes (T2D) and cardiovascular disease (CVD), imposing unprecedented burden on public health worldwide. MetDs stem from a complex interplay of multiple genes and cumulative exposure to environmental risk factors, yet the exact etiology remains elusive. To address this challenge, I embarked interdisciplinary systems biology studies encompassing the development of a multi-omics integration tool, elucidation of genetically perturbed tissue networks shared by T2D and CVD, and examination of environmentally perturbed gene networks by a prevalent endocrine disrupting chemical (EDC). First, I developed a multi-omics integration pipeline named Mergeomics, which consists of independent modules that 1) leverage multi-omics association data to identify biological processes that are perturbed in disease, and 2) overlay the disease-associated processes onto molecular interaction networks to pinpoint hubs as potential key regulators. Unlike existing tools that are mostly dedicated to specific data type or settings, the Mergeomics pipeline accepts and integrates datasets across platforms, data types, and species. The performance of Mergeomics was demonstrated by both simulation and case studies that include genome-wide, epigenome-wide, and transcriptome-wide datasets of total cholesterol and fasting glucose. I then applied Mergeomics to identify the shared gene networks between CVD and T2D through a comprehensive integrative analysis driven by five multi-ethnic genome-wide association studies (GWAS) for CVD and T2D, expression quantitative trait loci (eQTLs), ENCODE, and tissue-specific gene network models from CVD and T2D relevant tissues. The shared networks captured both known and novel processes underlying CVD and T2D. I also predicted 15 key drivers for the shared gene networks and cross-validated the regulatory role of top key drivers using in vitro siRNA knockdown, in vivo gene knockout, and two Hybrid Mouse Diversity Panels each comprised of >100 strains. Lastly, I leveraged systems biology approaches to assess the target tissues, molecular pathways, and gene regulatory networks associated with a developmental exposure to the model EDC Bisphenol A (BPA). Prenatal BPA exposure was found to cause transcriptomic and methylomic alterations in the adipose, hypothalamus, and liver tissues in mouse offspring, with cross-tissue perturbations in lipid metabolism as well as tissue-specific alterations in histone subunits, glucose metabolism and extracellular matrix. Network modeling prioritized main molecular targets of BPA, including Pparg, Hnf4a, Esr1, and Fasn. Moreover, integrative analyses identified the association of BPA molecular signatures with MetDs phenotypes in mouse and human. In summary, I presented the community a flexible and robust computational pipeline for multidimensional data integration, and offered mechanistic insights into the genetic and environmental underpinnings of MetDs by exploiting the power of systems biology through both computational and experimental approaches.

Physiological functioning of an organ system such as healthy brain functioning is enabled by orchestrated information flow between molecular entities within cells, between cells, and across tissues and its understanding requires a systems approach. In complex disease such as Alzheimer’s disease, there are numerous subtle, moderate, and strong perturbations across biological networks which result in disease onset and progression. High-throughput multi-omics profiling of genetic variants, the epigenome, transcriptome, proteome, metabolome, and microbiome is a first step in probing the complexity of biological systems and disease etiologies. There remains the problems of multi-omics integration, modeling and interpretation, and accessibility of computational analyses. To address these issues, in the methods development section of my dissertation research, I developed the complete overhaul of the web server for Mergeomics, a method for pathway- and network-level understanding of complex disease using multi-omics integration, which significantly improved the usability and functionality of the tool. To facilitate omics-driven drug discovery, I implemented the PharmOmics web server and the direct pipeline from Mergeomics, allowing users to gain immediate therapeutic insight from disease-associated pathways, key drivers, and key driver subnetworks. The second focus of my dissertation was the application of systems biology analytical tools to understand metabolic contributions to Alzheimer’s disease. I determined hippocampal and hypothalamic cell type specific alterations in response to amyloid-beta accumulation in the 5XFAD mouse model and investigated the effect of metabolically challenging 5XFAD mice with fructose overconsumption and supplementation with docosahexaenoic acid and nicotinamide riboside. I identified cell type- and tissue-resolved pathway and network mechanisms of disease exacerbation by metabolic challenge and disease alleviation by docosahexaenoic acid and nicotinamide. My efforts in developing the web server implementations of Mergeomics and PharmOmics have promoted greater accessibility of deriving mechanistic insight from big biological data for the broader scientific community, and use of these tools have allowed unbiased examination into the connection between metabolic syndrome and Alzheimer’s disease and uncovered potential therapeutic targets and strategies via dietary modulation.

Spontaneous coronary artery dissection (SCAD) and coronary artery disease (CAD) both represent cardiovascular diseases that may result in myocardial infarction and sudden death. However, the populations predominantly impacted by SCAD and CAD as well as the pathogenesis of both diseases are notably different; while SCAD overwhelmingly affects a young, female population and is non-atherosclerotic, CAD mainly impacts males and develops via atherosclerosis. The genetic architecture and mechanisms leading to SCAD onset are currently poorly understood, and although the genetic and environmental risk factors for CAD are better elucidated, further understanding of the mechanisms in which genetic and environmental factors interact to facilitate pathogenesis is needed. We conducted two studies to examine the networks of SCAD and CAD at various omics layers. In a comprehensive multiomics human SCAD study, we investigated disease-associated biological mechanisms and modeled tissue-specific gene regulatory and protein-protein interaction networks across genetic, transcriptomic, and proteomic levels and predicted potential therapeutics for SCAD treatment. We identified various pathways related to the extracellular matrix, immune function, and blood clotting activation. We pinpointed key regulatory genes and hub proteins central within our tissue-specific networks, including HOXB9, CADM1, and COL18A1. Female-specific drugs, such as medroxyprogesterone and progesterone receptor agonist, were prioritized in SCAD drug repositioning. In a single-cell RNA-sequencing mouse study, we examined the cell type-specific transcriptome changes of the aorta in Ldlr-/- mice, an established atherosclerosis animal model for CAD, under various diet conditions, namely chow, high-cholesterol, and high-cholesterol with added trimethylamine N-oxide (TMAO). We identified significant differences in cellular composition in the modulated smooth muscle cell (SMC) cluster and macrophage M1 subtype and elucidated the transition of SMCs into a protective fibromyocyte phenotype in a data-driven manner. Additionally, we determined cell type-specific and shared differentially expressed genes and their corresponding biological pathways across dietary conditions, pinpointing extracellular structure organization pathways as significant in various cell types between high-cholesterol and chow diets as well as apoptosis-related pathways as enriched in SMC and modulated SMC with the addition of TMAO to the high-cholesterol diet. We further predicted intercellular communications that influence downstream gene expression in modulated SMC and endothelial cells.

Complex diseases such as Alzheimer’s disease are driven by molecular changes in many cell types in different tissues. Recent advances in scRNAseq and spatial transcriptomics provide tools to determine cell type specific effects in individual tissues resulting from genetic and environmental perturbations. Properly interpreting these data require computational tools and biologically rooted analyses to identify key mechanisms underlying complex diseases. Here we design, develop, and apply computational methods for integrating scRNAseq and spatial transcriptomics data to identify mechanisms underlying pathogenesis of disease and potential therapeutics. First, we designed a deep learning approach, JSTA, for integrating scRNAseq and spatial transcriptome data from multiplexed FISH for cell segmentation and cell type annotation, revealing spatially distributed cell subtypes and spatially differentially expressed genes in the mouse hippocampus. Next, we developed a gradient-boosting machine based approach, SCING, for identifying cell type specific gene regulatory networks (GRN) using scRNAseq and spatial transcriptomics data. This tool provides GRN subnetworks annotated with biological pathways for associating subnetwork expression with disease phenotypes and spatial domains. We applied these and other existing tools to scRNAseq and spatial transcriptomics datasets to understand the mechanism underlying diverse types of diseases or physiological traits, including Alzheimer’s disease and heart innervating neurons and satellite glial cells in the stellate ganglion in the context of dilated cardiomyopathy. Our studies established new computational tools applicable to diverse types of single cell omics data and revealed biological insights to complex diseases.

Type 1 diabetes (T1D) is a complex disease, involving a genetic predisposition that interacts with environmental triggers, leading to the loss of insulin producing beta cells in the pancreas. However, the molecular cascades underlying T1D are poorly understood and remain to be explored. We hypothesize that genetic risk factors of T1D perturb tissue-specific biological pathways and gene networks, which ultimately leads to the pathogenic end point in beta cells. We sought to identify the gene networks and key regulators for T1D by conducting a comprehensive, data-driven multi-omics analysis that integrates human genome-wide association studies (GWAS) of T1D, tissue-specific genetic regulation of gene expression in the form of expression quantitative trait loci (eQTLs), and tissue-specific gene network models using a computational pipeline Mergeomics. Our integrative genomics approach revealed immune pathways such as adaptive immune system, cytokines and inflammatory response, ZAP70 translocation, primary immunodeficiency and immunoregulatory interactions between a lymphoid and non-lymphoid cell, across various tissues. We also identified tissue-specific signals such as regulation of complement cascade in adipose tissue, macrophages, and monocytes, NOTCH signaling in adipose tissue and macrophages, protein folding, calcium signaling chemotaxis and lysosomal pathways in the pancreas, adipose, and monocytes, and viral infection in macrophages and monocytes. Network modelling of these pathways highlights a number of key regulator genes such as GBP1, USP18, STAT1, RPL17, HLA genes (HLA-A,-B,-C, and –G), and immunomodulatory genes (LCK, VAV1, ZAP-70), each of which has suggestive roles in the pathophysiology of T1D or other autoimmune disorders. Together, our integrative genomics approach offers comprehensive insights into the tissue-specific molecular networks and regulators as well potential between-tissue interactions underlying T1D, with potential for guiding future development of therapeutic strategies targeting the disease.

Common complex metabolic diseases (MetDs) such as obesity, type 2 diabetes (T2D), coronary artery disease (CAD) and non-alcoholic fatty liver disease (NAFLD), impose an unprecedented burden on public health worldwide and demonstrate sex differences. Our general hypothesis is that genetic risk factors perturb set of genes in the form of functional gene networks, which subsequently induces the initiation and progression of MetDs. Following this hypothesis, our research focuses on dissecting the molecular networks that are perturbed by genetic risk factors of MetDs utilizing multiomics systems biology approaches. To address this challenge, I embarked interdisciplinary systems biology studies encompassing the development of an accessible multi-omics integration webserver, elucidation of genetically perturbed tissue networks in numerous MetDs, and uncovering the relative contribution of three sex factors in gene regulation in tissues relevant to MetDs. First, I contributed to the development of a user-friendly webserver for multiomics integration, network modeling, and network-based drug repositioning for complex diseases such as MetDs. Second, I investigated the genetically perturbed gene networks that underly various MetDs, namely, lipid traits, diabetes, CAD, and NAFLD. Third, I employed systems biology approaches to uncover the individual and interactive contribution of three sex factors (sex chromosomes, gonads, and sex hormones) in gene regulation in tissues relevant to MetDs. Completion of these projects offer a user-friendly bioinformatic tool, molecular insights, and drug candidates for MetDs.

The genetic heterogeneity of autism spectrum disorder (ASD) has been a long-standing obstacle in our understanding of the pathogenic mechanisms of the disease, as the genetic risk of ASD is made up of numerous common variants and rare de novo or inherited variants. Previous studies have focused primarily on identifying rare variants and their impact on brain cortical cell types, and these mutations have been found to primarily affect neurodevelopment by perturbing neuronal functions. By contrast, common variants have been found to contribute substantially to ASD heritability, but remain understudied. This suggests a need to consider both rare and common variants of ASD to understand the genetic mechanisms of the disease. Furthermore, previous studies have implicated the subcortical areas of the brain and other organ and tissue systems such as the digestive and immune systems in ASD, but tissue-specific mechanisms remain poorly explored. To address these knowledge gaps, this thesis aims to identify gene networks and pathways informed by ASD common variants in both brain and peripheral tissues across the body and further examine whether these networks also capture genes informed by rare variants. We achieve this by integrating tissue level RNA sequencing data, genome wide association study (GWAS) summary statistics, and tissue-specific transcriptional regulatory networks using the multiomics integration method Mergeomics. Furthermore, we infer tissue-specific key regulatory genes governing the pathways and networks of ASD common variants by leveraging tissue-specific Bayesian gene regulatory networks. Lastly, we investigate whether the gene networks informed by ASD common variants converge with those of known ASD rare variants. Our multitissue multiomics systems studies incorporating both common and rare variants reveal the key tissues, biological pathways, and gene network regulators of ASD and identify key similarities and differences between ASD common and rare variants in tissue and network specificity.