Microorganisms residing in the human body are more than nutrient eaters. The intimate relationship between them and us contributes to human health and disease in myriad ways. Previous developments in genome-scale microbial metabolic modeling have demonstrated its power to uncover metabolic potential, but data availability limits the species and strains that can be covered. To expand on their earlier work by capitalizing on data with a much broader sample coverage, Ines Thiele from the University of Galway and colleagues present AGORA2, a resource of >7,000 metabolic reconstructions of human microorganism strains.
While the basic theoretical principles of metabolic reconstruction have long been explored, building a framework of high-efficiency and high-quality data collection, curation, model generation and quality control was the key in this endeavor, considering the huge diversity in microbial metabolism. “Over the past decade many great reconstructions pipelines have been built automating many reconstruction steps. However, the manual refinement against existing experimental data was often not possible or not easy,” says Thiele, “We set out to also automate this step. The collection of experimental data from the literature remains mostly a manual step, but the propagation to strains for a species with experimental data is now automated.” Other challenges they overcame include gap filling and debugging, as well as optimization and parallelization of the code, notes Thiele. “While one optimization problem can be solved in less of a second, performing time-consuming tasks such as flux variability analysis (where each reactions in a model is maximized and minimized) across 7,000 reconstructions needed new approaches and efficient implementations.”
This is a preview of subscription content, access via your institution