Microbiome Immunity Project Researchers Create Ambitious Plans for Data


The Microbiome Immunity Project researchers—from Boston, New York, and San Diego—met in person a few weeks ago to make plans that include a 3D map of the protein universe and other far-ranging uses for the data from the project.




The research team members pictured above are (from left to right): Vladimir Gligorijevic (Simons Foundation’s Flatiron Institute), Tommi Vatanen (Broad Institute of MIT and Harvard), Tomasz Kosciolek (University of California San Diego), Rob Knight (University of California San Diego), Rich Bonneau (Simons Foundation’s Flatiron Institute), Doug Renfrew (Simons Foundation’s Flatiron Institute), Bryn Taylor (University of California San Diego), Julia Koehler Leman (Simons Foundation’s Flatiron Institute). Visit the project's Research Participants page for additional team members.

During the week of May 28, researchers from all Microbiome Immunity Project (MIP) institutions (University of California San Diego, Broad Institute of MIT and Harvard, and the Simons Foundation’s Flatiron Institute) met in San Diego to discuss updates on the project and plan future work.

Our technical discussions included a complete overview of the practical aspects of the project, including data preparation, pre-processing, grid computations, and post-processing on our machines.

We were excited to notice that if we keep the current momentum of producing new structures for the project, we will double the universe of known protein structures (compared to the Protein Data Bank) by mid-2019! We also planned how to extract the most useful information from our data, store it effectively for future use, and extend our exploration strategies.

We outlined three major areas we want to focus on over the next six months.

  • Structure-Aided Function Predictions

We can use the structures of proteins to gain insight into protein function—or what the proteins actually do. Building on research from MIP co-principal investigator Richard Bonneau’s lab, we will extend their state-of-the-art algorithms to predict protein function using structural models generated through MIP. Using this new methodology based on deep learning, akin to the artificial intelligence algorithms of IBM, we hope to see improvements over more simplistic methods and provide interesting examples from the microbiome (e.g., discover new genes creating antibiotic resistance).

  • Map of the Protein Universe

Together we produce hundreds of high-quality protein models every month! To help researchers navigate this ever-growing space, we need to put them into perspective of what we already know about protein structures and create a 3D map of the “protein universe.” This map will illustrate how the MIP has eliminated the “dark matter” from this space one structure at a time. It will also be made available as a resource for other researchers to explore interactively.

  • Structural and Functional Landscape of the Human Gut Microbiome

We want to show what is currently known about the gut microbiome in terms of functional annotations and how our function prediction methods can help us bridge the gap in understanding of gene functions. Specifically, we want to follow up with examples from early childhood microbiome cohorts (relevant to Type-1 diabetes, or T1D) and discuss how our methodology can help us to better understand T1D and inflammatory bowel disease.

The future of the Microbiome Immunity Project is really exciting, thanks to everyone who makes our research possible. Together we are making meaningful contributions to not one, but many scientific problems!


Related Articles