Machine learning is starting to understand the human gut

The communities formed by human gut microbes can now be predicted more accurately with a new computer model developed in a collaboration between biologists and engineers led by the University of Michigan and the University of Wisconsin.

Creating the model also suggests a route to scaling up from the 25 microbe species studied up to the thousands that may be present in the human digestive system.

“Every time we increase the number of species, we get an exponential increase in the number of possible communities,” said Alfred Hero, the John H. Holland Distinguished University Professor of Electrical Engineering and Computer Science at the University of Michigan and co-corresponding author of the study in the journal eLife†

“That’s why it’s so important that we can extrapolate from the data collected on a few hundred communities to predict the behavior of the millions of communities we haven’t seen.”

As research continues to reveal the multifaceted ways in which microbial communities affect human health, probiotics often fail to live up to the hype. We don’t have a good way of predicting how the introduction of one species will affect the existing community. But machine learning, an artificial intelligence approach where algorithms learn to make predictions based on data sets, could change that.

“Problems of this magnitude required a complete overhaul of how we model community behavior,” said Mayank Baranwal, an adjunct professor of systems and control engineering at the Indian Institute of Technology, Bombay, and co-first author of the study.

He explained that the new algorithm can map the entire landscape of 33 million possible communities in minutes, compared to the days to months required for conventional ecological models.

Microbial Sim Cities

Integral to this important step was Ophelia Venturelli, assistant professor of biochemistry at the University of Wisconsin and co-corresponding author of the study. Venturelli’s lab conducts experiments on microbial communities and maintains them in low-oxygen environments that mimic the mammalian gut environment.

Her team created hundreds of different communities with microbes that occur in the human colon and the healthy state of the gut microbiome† They then measured how these communities evolved over time and the concentrations of key health-relevant metabolites, or chemicals, produced when the microbes break down food.

“Metabolites are produced in very high concentrations in the gut,” Venturelli said. “Some are beneficial to the host, such as butyrate. Others have more complex interactions with the host and the gut community.”

The machine learning model allowed the team to design communities with the desired metabolite profiles. This kind of monitoring could eventually help doctors find ways to treat or protect against disease by introducing the right microbes.

Feedback for faster model building

While human gut microbiome research has a long way to go before it can provide these types of interventions, the approach the team developed could help get there faster. Machine learning algorithms are often produced using a two-step process: collecting the training data and then training the algorithm. But the feedback step added by the team at Hero and Venturelli provides a template for quickly improving future models.

Hero’s team initially trained the machine learning algorithm on an existing dataset from the Venturelli lab. The team then used the algorithm to predict the evolution and metabolite profiles of new communities that Venturelli’s team built and tested in the lab. While the model performed very well overall, some of the predictions identified weaknesses in model performance, which Venturelli’s team backed up with a second round of experiments, closing the feedback loop.

“This new modeling approach, coupled with the speed with which we could test new communities in the Venturelli lab, would enable the design of useful microbial communities,” said Ryan Clark, co-first author of the study, who was a postdoctoral researcher. in Venturelli’s lab when he did the microbial experiments. “It was much easier to optimize for the production of multiple metabolites at once.”

The group chose a long-term memory neural network for the machine learning algorithm, which accounts for sequence prediction problems. However, like most machine learning models, the model itself is a ‘black box’. To find out which factors were factored into its predictions, the team used the mathematical map produced by the trained algorithm. It revealed how each type of microbe affected the abundance of the others and what types of microbes metabolites it supported. They could then use these relationships to design communities worth exploring through the model and in follow-up experiments.

The model can also be applied to a variety of microbial communities outside of medicine, including accelerating the breakdown of plastics and other materials for environmental clean-up, producing valuable compounds for bioenergy applications, or enhancing plant growth.