PhD project: Designing a new tool to find disease-associated genes

One of my PhD projects involved developing and testing a new bioinformatics algorithm (which we called ‘GeneSurrounder’) to find individual genes that influence disease associated mechanisms. This project was joint work with Dr. Rosemary Braun.

We recently presented a poster of our work at the ISMB 2018 conference. The work is currently under review and available on the arXiv.

Scientific summary: Network-based identification of disease genes in expression data

New technologies have made it possible to probe complex diseases and identify genes for precision medicine. These technologies yield large datasets that require advanced algorithms to extract useful information. Since genes interact with one another, we seek specific genes that have an effect on the whole system. However, existing tools either neglect interactions between genes, or yield “pathway-level” results that are difficult to target. There is therefore a need for algorithms that can precisely identify genes for treatment while taking into account the interaction network.

To fill this gap, we present GeneSurrounder, a new algorithm that ranks genes based on the evidence that they are sources of disruption on the network of interacting genes. Since the effects of a “disruptive” source gene would propagate outward in the interaction network, we find these genes by searching for a telltale pattern of attenuating and correlated biological signal in the data. We apply GeneSurrounder to three distinct ovarian cancer datasets and demonstrate that the results are more reproducible than competing techniques. We also find that our method is able to identify genes known to impact ovarian cancer. These results suggest that GeneSurrounder is able to reproducibly detect specific therapeutic gene targets.