Genome-wide association studies (GWAS) have identified hundreds of thousands of associations between genetic variants and human complex traits/diseases. To functionally annotate the trait/disease-associated variants, extensive efforts are made to study the genetic effects on downstream molecular phenotypes in a wide variety of tissue types and cell types. Genetic effects on functionally related `omic' traits often co-occur in relevant cellular contexts, such as tissues. In Chapter 2, motivated by the multi-tissue methylation quantitative trait loci (mQTLs) and expression QTLs (eQTLs) analysis of Genotype-Tissue Expression project, we propose X-ING (Cross-INtegrative Genomics) for cross-omics and cross-context integrative analysis. A major innovation of the method is that it models latent association indicators instead of effect sizes and uses multi-view learning to account for major patterns among latent indicators across omics data types and tissue types. This facilitates integrative analysis of different data types of different effect distributions. Moving beyond the integrative association analysis, in Chapter 3 we develop a multi-context multivariable integrative Mendelian randomization framework, mintMR, for mapping expression and molecular traits as joint exposures. The proposed method overcomes the unique challenges in mapping risk genes, and these challenges are under-addressed by conventional Mendelian randomization methods. MintMR improves the estimation of sparse tissue-specific causal effects of multiple genes with a limited number of IVs by simultaneously modeling the latent tissue indicators of disease relevance across multiple gene regions and subsequently improving the estimation of latent disease-relevant probabilities. In Chapter 4, we further expand the framework to study risk genes in specific cell types using deep learning methods. Single-cell RNA sequencing (scRNA-seq) enables the high-throughput profiling of gene expression specific to cell types. We proposed a deep-cellMR method capturing the nonlinear and complex dependencies across cell types and further improving the estimation of cell-type-specific effect of each gene in each cell. The proposed methods in this dissertation can be broadly applied to map multi-omics QTLs and study risk genes for complex traits and diseases, and they can be applied to many other data types for conducting integrative association and causal analyses.