Obesity and obesity-related diseases have become an increasing burden to the health systems worldwide. Since the obesity-related diseases, such as type 2 diabetes (T2D), non-alcoholic fatty liver disease (NAFLD), and cardiovascular disease, share many risk factors and affect multiple human tissues, the complexity of these diseases has prevented disentangling the underlying causal effects. To investigate the cross talk between various tissues and obesity-related complex diseases, we comprehensively analyzed single nucleus RNA sequence (sn-RNA-seq) data, bulk RNA-seq data in multiple tissue types, as well as genotype data in several independent cohorts. First, in Chapter III we utilized the adipose sn-RNA-seq data as a reference to estimate the cell-type composition in human subcutaneous adipose tissue. Using body mass index (BMI), adipose mitochondrial (MT) gene expression, and the estimated cell-type proportions as predictors, we explained ~40% of variance in systemic insulin resistance and accurately estimated insulin resistance in human cohorts with adipose RNA-seq data. Our iianalysis discovered the important role of adipose transcriptional activity and MT activity in the development of systemic insulin resistance. Moreover, in Chapter IV we developed another prediction model that utilized BMI, waist circumference, age, sex, and serum lipid, liver enzyme, and glucose levels to accurately predict non-alcoholic fatty liver (NAFLD) in the UK Biobank (UKB) cohort. The novel NAFLD score (NAFLDS) model achieved a high accuracy (AUC = 0.9) and outperformed the existing fatty liver index (FLI) in predicting the NAFLD status in UKB. Using NAFLDS as the surrogate of the NAFLD status, we utilized cis expression quantitative trait loci (cis-eQTLs) in liver and coronary arteries to refine the instrumental variables (IV) for our Mendelian randomization (MR) analyses and demonstrated a one-way causal effect of NAFLD on CAD (beta = 0.024, p-value = 9.4e-6). While analyzing the RNA-seq cohorts, we found that allele-specific expression (ASE) analysis is a powerful tool that improves the accuracy and power in identifying cis gene regulation in RNA-seq cohorts. However, the reference alignment bias remains a major obstacle in ASE analysis. The existing methods are either inaccurate or relatively slow, which makes it impractical to accurately estimate ASE events in larger cohorts. To address this issue, we developed ASElux that uses personal genotype data as the reference to fast and accurately align the ASE reads to both alleles. By applying ASElux into the GTEx lung samples, we showed that ASElux is at least ~5X faster than any existing method, while achieving a top accuracy. In summary, we developed a novel method ASElux to resolve the reference alignment bias in the ASE analysis. Furthermore, we comprehensively combined multi-omics data from adipose, liver, and coronary artery tissues and established two causal effects regarding the obesity related diseases (obesity -> insulin resistance/pre-diabetes and NAFLD -> CAD).