The advent and continual progression of technologies for profiling DNA variation, RNA transcription, and protein abundance have fueled the respective fields of genomics, transcriptomics and proteomics. These parallel advancements pave the way for integrative modeling of multiple, heterogeneous data sources, with the important goal of providing holistic models of cellular physiology. Indeed, as these rich sources of information provide immense potential to model biological processes and learn the genetic architectures for complex phenotypes, the onus falls on computational scientists to develop methodologies that provide hypotheses to guide experimentation.

In this dissertation, we present three computational methodologies that integrate multiple data sources for the following purposes: predicting synthetic lethal interactions between two genes, causal network reconstruction, and association mapping by sampling mixed models.