Wednesday Colloquium: Professor Rosemary Braun, Northwestern University  Add To Calendar

  • Date(s): Wednesday, 4/16 3:00 PM to Wednesday, 4/16 4:00 PM
  • Speaker: Professor Rosemary Braun, Northwestern University
  • Host:
  • Campus Address: 238 SES Science and Engineering South
Title: Hearing the Shape of Life: Spectral Analysis of Gene Regulatory Networks

Abstract: Modern high-throughput assays, which yield highly detailed genomic profiles of each sample, now provide an immense opportunity to reveal the genes and regulatory mechanisms underlying disease. At the same time, the complexity of these diseases poses a challenge: a crucial mechanism may be affected in diverse ways even amongst phenotypically similar samples, and the adaptive robustness of living processes can enable tolerance at the systems level to differences at the gene level. There is thus a need for analytical techniques that go beyond gene-level association statistics to characterize systems-level differences in regulatory networks (pathways). 

In this talk, I will discuss a novel graph-theoretical approach for identifying altered gene regulatory networks from high-throughput data. Just as the Laplacian of a physical system (eg, that of a drumhead) can be used to infer its dynamical properties (its sound), spectral decomposition of the graph Laplacian provides a means by which the dynamical properties of a network may be summarized. By constructing weighted network models from genomic data and computing the graph Laplacian, we encapsulate bulk pathway-wide patterns of gene expression across the regulatory network. By comparing the spectra between phenotypes, we can identify globally-altered pathways with high accuracy without requiring their alterations to be caused by the same gene in each sample, and also allowing for the possibility of compensatory changes that lead to large differences at the gene-level yet preserve the network connectivity. Analysis of the associated eigenvectors reveals the influence of individual genes on the behavior of the pathway, and enables us to make testable predictions about the dynamics of the system in response to perturbations. I will describe the method in detail and demonstrate its application to cancer data.