Precise functions of genes in multi-cellular organisms such as humans are defined by the context of tissues where the genes are expressed. However, the variability and context-specificity of gene function across different tissues are not well-understood yet because, high-throughput interaction measurements are largely infeasible in solid tissues and low-throughput experiments are highly skewed towards well studied genes. Recently, researchers have started characterizing context-specific gene representations in a data-driven manner with some success. However the representations remain noisy and comprehensive analysis of tissue-specific gene functions in complex human diseases remains untouched. In this project, we are developing a new data-integration framework that finds more robust tissue-specific representations of genes and their functional relationships leveraging the topological properties of genes in tissue-specific as well as global interactomes. By decoupling the dimensionality of feature representations from genes, we can achieve substantial power in the functional inference of context-specific gene interactions. We are using the tissue-specific functional representations of genes across tissues to derive a map of tissue-specific disease relationships. This can be immensely helpful in understanding complex diseases like cancer, which originate in one tissue and then can metastasize to many other tissues, or autism spectrum disorder which is primarily a neurodevelopmental disorder but is often accompanied by a wide range of other conditions manifested in multiple organ systems.