Speaker: Eric Schadt, Dept. of Genetics, Rosetta Inpharmatics, LLC/Merck Research Labs
When: Mon, Feb 11, 2008
Where: CNSI Auditorium
Abstract:
A primary aim in systems biology research is the construction of networks that predict complex system behavior like disease. Complex phenotypes like common human diseases can result from complex network interactions within and between tissues, where the network states are a function of genetic background and environmental factors. Most human genetic studies have as a primary aim the identification of changes in DNA that correlate with changes in disease traits. An alternative to this approach involves the reconstruction of molecular networks that actually define the disease, where such networks are constructed by integrating genetic, gene expression, transcription factor binding site, protein interaction, and other molecular phenotype data across multiple tissues relevant to the disease under study. I present a novel integrative genomics method that combines multiple types of large-scale molecular data, including genotypic, gene expression, TFBS, and PPI data to construct causal, probabilistic gene networks. I demonstrate the importance of incorporating systematic sources of perturbations to infer causal relationships among genes by reconstructing whole gene networks based on different types and subsets of data. I also demonstrate that networks resulting from this integrative approach enable the direct identification of genes causal for disease, beyond what could be achieved by a classic human genetic approach. I demonstrate that the predictive networks resulting from this process provide a far richer context within which to interpret the functioning of any given gene, compared to single gene approaches, which in turn leads to efficient strategies for prioritizing nodes in the network for therapeutic intervention. In this way, nodes predicted to have a larger impact on disease associated traits can be identified and pursued over genes predicted to have lesser effects. Finally, I demonstrate how cross-tissue networks can also be constructed in this way, and that such networks elucidate coherent pathways associated with disease that are not visible by examining single tissues alone, providing yet another level of complexity that can be leveraged to elucidate the most efficient nodes in the network to target to achieve the largest impact on disease. Several examples of genes identified in this way that are under active development are discussed.