A long-term project that illustrates both the impact of our work and our typical development process is the collaboration with research biologists at the Feinberg School of Medicine at Northwestern University, and at the University of Illinois.

Our objective is to identify relevant features for diagnosing and understanding a disease termed Interstitial Cystitis/Bladder Pain Syndrome (IC/BPS). This disease is associated with debilitating pelvic pain and lacks diagnostic biomarkers and effective therapies. The data includes genetic sequences of gut microbes and metabolites of healthy and IC/BPS patients. To accomplish our objective, we modify computationally intensive, multivariate machine-learning approaches that have shown promising results in other scenarios. Our initial approach used the Random Forest algorithm followed by stability analysis and feature permutation to determine and rank all relevant features.

Results of this analysis showed that particular microbial species are relevant to modulating pain signals. Following our analysis, more target sequencing and several mouse studies were conducted. This work resulted in a Nature publication, multiple conference presentations, a patent for a microbiome-related disease biomarker and an NIH grant to sufficiently support our two teams on this project for five years.