The end goal of medical genomics research is to predict, prevent, and treat disease. With our research, we are developing computational genomics methods aimed at the prediction component. While we are interested in a number of diseases and systems, our primary effort focuses on lung disease and physiology as a model system. We are working in collaboration with Dr. Ronald G. Crystal's group for this research. The Crystal group has a system in place for collecting high-throughput genomic data for relevant lung tissues and their knowledge of the lung system provides critical insight into where computational genomic methodologies can have the greatest impact. Our current research includes development of methods for analyzing next generation sequencing data to characterize gene expression profiles in the lungs of both healthy people and individuals with disease, feature selection methods aimed at developing accurate biomarkers, and application of our genome-wide association and network discovery methodologies to the lung system.


Example of how smoking modulates the expression of genes modulated by Nrf2 (an oxidant-responsive transcription factor), where green and red reflect down- and up-regulation respectively (left). The histograms present the predictive ability of an index (biomarker) based on these genes, where colored bars reflect quantiles of the index (top right) and variability (bottom right), when compared to an index based on other gene groups (Hubner et al. 2009)
Subspace (top) and individual genes (bottom) extracted from a Higher Order Singular Value Decomposition (HOSVD) of genome-wide gene expression profiles of the small airway epithelium and trachea of the lung that captures the chief differences between smokers and non-smokers. Methods of this type are being used to develop biomarkers and to assess cases when the profiles of certain tissues are sufficiently similar to important cell types (that are difficult to collect) such that they can be used to generate “stand-in” or “canary” profiles. From Omberg et al. in prep.