Gut Feeling

Posted on Tuesday, September 8th, 2020

Woman with hands outstretched, cradling images of various colourful bacteria
Distinct combinations of gut microbial communities found in different individuals are called enterotypes. There are three main enterotypes in humans, which are linked to diet.

U of G researchers develop a mathematical model to examine patterns in the human gut microbiome.

Trillions of microorganisms live inside the human body, forming an ecosystem that can impact our physiology. Bacteria, viruses, and fungi are examples of microbiota, and in humans, most of them reside in the gut. Together, microbiota form a community of species called a microbiome. Understanding the impact of the gut microbiome on human health and disease is a popular area of study. A healthy microbiome can aid in digestion, while an unhealthy microbiome has been linked to diseases like cancer. A key challenge for researchers, however, lies in how to assess what combination of microbiota determines a “good” versus “bad” human gut microbiome, while accounting for external biological and environmental influences, such as individual diet or drug intake.

University of Guelph mathematics and statistics professor Zeny Feng and her research team are working to address that challenge through mathematical modelling. The researchers have developed a new model for investigating patterns in human gut microbiomes, which will enable researchers to assess how these patterns are linked to human health. The U of G team has enhanced a mathematical technique called a cluster analysis. Cluster analysis is an exploratory method that groups a set of data into “clusters” based on similarities. Applied to the human microbiome, these methods involve examining individual variation in the human gut microbial ecosystem and using machine learning and mathematical algorithms to classify individuals with similarities into groups. Grouping individuals can have clinical benefits. For example, clinicians can assess microbiome differences between healthy and diseased individuals to inform recommendations.

The U of G researchers proposed a model-based clustering approach, and isolated the best model among a few based on a series of simulations. They identified the best-fitting models and applied them to a real dataset that consisted of healthy individuals and individuals with colon cancer.

“To understand the effects of the human microbiome on health, it is crucial to understand the underlying structure of microbial communities in the human gut. Our method serves to group individual microbiomes together based on their composition, while disentangling the effects that biological and environmental factors may have on these clusters.”

Subedi S, Neish D, Bak S, Feng Z. Cluster analysis of microbiome data by using mixtures of Dirichlet-multinomial regression models. J R Stat Soc C-Appl. 2020 Jul 26. doi: 10.1111/rssc.12432.

News Archive