Blue Waters research integrates data to predict E. coli behavior

by Susan Szuch

Before you savor the first bite of that delicious snack you have tucked in your bag to get you through the mid-afternoon slump, take a moment to thank Ilias Tagkopoulos—it’s researchers like him that help keep your food safe from nasty bacteria.

A computer scientist by trade, Tagkopoulos runs an experimental microbiology lab at University of California-Davis. He has been using the Blue Waters supercomputer to predict the cellular behavior of E. coli. The findings were recently published in Nature Communications.

“The level of integration in advanced machine learning, high performance computing and experimental microbiology is something rare,” he says. “It’s really an interdisciplinary project that could not be done by just one discipline.”

While anyone can get sick from an E. coli infection, it can be life threatening to young children and elderly people. Most recently, the fast food restaurant chain Chipotle has been trying to regain customer trust after a major outbreak of the bacteria, which can cause severe illness including stomach cramps, diarrhea and vomiting.

Similarly, the rise of antibiotic resistance limits the effectiveness of current antibiotics and makes us vulnerable to microbial infections. Tagkopoulos and his lab works together with academic and industrial partners to finds ways to predict how these behaviors emerge and how to control them.

Research on how organisms express genes and proteins has been conducted for decades, but each lab had their own methods, making it difficult for researchers to compare findings. While studies have been done before to standardize and compile information, Tagkopoulos’ lab has produced the most comprehensive work so far.

Tagkopoulos’ lab compiled already-existing data and then integrated it, so that it would be possible to compare and analyze findings. Once this was done, it became the “perfect data set for integrative modeling,” allowing the researchers to create models for each layer of the E. coli organism. Like pieces of a puzzle, those models came together to create a model of the whole organism.

“Blue Waters is absolutely instrumental here because of the size of the simulations. We needed computational power,” Tagkopoulos says. “If we didn’t have Blue Waters, this level of integration and learning would be very difficult.”

The simulations run on Blue Waters also provide other researchers with more options for predicting organism behavior.

“The model can predict the cellular state for novel conditions and provide a confidence measure for those predictions. This can be quite useful to researchers to have a fast and inexpensive way on how the organism might behave in a specific experimental setting,” Tagkopoulos says. “Applications range from finding the most likely conditions and strains for maximizing protein production and growth to identifying key pathways for antibiotic and stress resistance.”

After E. coli, Tagkopoulos is extending the research to Salmonella, with the next compendia being ready around the summer of 2017.

“Given its prevalence in food poisoning, having a resource like a predictive model of how Salmonella will behave in a certain environment will be beneficial from a food safety perspective,” Tagkopoulos says.

The research being done on Blue Waters has potential applications for biotechnology fields, as well, in extending to research on the bacterium Bacillus subtilis, one of the workhorses in the field of biotechnology.

In spite of the strides they’ve made in research and creating advanced models, there’s still a ways to go before it can be applied in an industrial setting and accessible for non-experts to use.

“What we found out is that we have sampled only a small fraction of possible experimental space, and that with incomplete information and metadata,” Tagkopoulos says. “We need to extend the data collected to settings that are relevant for industrial problems, such as various fermentation conditions; fine-tune the normalization and quality assurance pipelines; and develop a better visualization interface for non-experts to be able to use the data, model and simulation resources.”

Disclaimer: Due to changes in website systems, we've adjusted archived content to fit the present-day site and the articles will not appear in their original published format. Formatting, header information, photographs and other illustrations are not available in archived articles.