Loading…
This event has ended. View the official site or create your own event → Check it out
This event has ended. Create your own
View analytic
Wednesday, August 2 • 15:30 - 15:50
Machine learning identifies unique taxa differentiating proximal and distal human colonic microbiota

Sign up or log in to save this to your schedule and see who's attending!

Colorectal cancer (CRC) remains a leading cause of death worldwide. Tumors of the proximal (right) and distal (left) colon are morphologically and genetically distinct. Previous work from our group found that microbial dysbiosis is associated with the development of colorectal cancer tumors in studies of both mice and humans. Analysis of the fecal microbiota from healthy and CRC patients further revealed different microbial signatures associated with disease. In this study, we extended our observations of the fecal microbiome to analysis of the proximal and distal human colon. We used a two-colonoscope approach on subjects that had not undergone standard bowel preparation procedure. This technique allowed us to characterize the native proximal and distal luminal and mucosal microbiome without prior chemical disruption. 16S rRNA gene sequencing was performed on proximal and distal mucosal biopsies, luminal and exit stool for 20 healthy individuals. Diversity analysis of each location revealed that each site contained a diverse community, and that a patient’s samples were more similar to each other than to that of other individuals. Since we could not differentiate sites along the colon based on community structure or community membership alone, we employed the machine-learning algorithm Random Forest to identify key species that distinguish biogeographical sites. Random Forest classification models were built using taxa abundance and sample location and revealed distinct populations that were found in each location. Peptoniphilus, Anaerococcus, Enterobacteraceae, Pseudomonas and Actinomyces were most likely to be found in mucosal samples versus luminal samples (AUC = 0.925). The classification model performed well (AUC = 0.912) when classifying mucosal samples into proximal or distal sides, but separating luminal samples from each side proved more challenging (AUC = 0.755). The left mucosa was found to have high populations of Finegoldia, Murdochiella and Porphyromonas. Proximal and distal luminal samples were comprised of many of the same taxa, likely reflecting the fact that stool moves along the colon from the proximal to distal end. Finally, comparison of all samples to fecal samples taken at exit uncovered that the feces were most similar to samples taken from the left lumen, again reflecting the anatomical structure of the colon. Taken together, our results have identified distinct bacterial populations distinct of the proximal and distal colon. Further investigation of these bacteria may elucidate if and how these groups contribute to differential oncogenesis processes on the respective sides of the colon.

Speakers
avatar for Kaitlin Flynn

Kaitlin Flynn

Postdoctoral Fellow, University of Michigan Medical School


Wednesday August 2, 2017 15:30 - 15:50
Graduate School of Management Building, room 309 Volkhovskiy Pereulok, 3, St. Petersburg, Russia