Modeling Microbial Communities with Metaflux

Modeling Microbial Communities with MetaFlux

Metabolic modeling has great potential for moving our understanding of the human microbiome to a predictive level. Predictive microbiome modeling could lead to breakthroughs in health care, such as the ability to rationally predict how to perturb a diseased microbiome to a healthy state. But metabolic modeling of the microbiome presents many challenges. This talk will survey some of these challenges, and will describe open research problems, potential solutions, and actual solu- tions available from our group’s Pathway Tools software. A common theme is that although new modeling algorithms may be important for microbiome modeling, many of the challenges we face are challenges of scale that must be met by providing better engineered modeling software.

Despite many years of research, building metabolic models is still technically challenging and can require 1–2 years of effort, even for experienced groups [7]. That number from 2010 may not reflect recent advances in modeling methodology. As a hypothetical example, if we assume two months per model, creating models for 300 key organisms within the human microbiome could cost $5M (assuming 50 person-years of effort, at $100K per person-year). That number assumes no duplication of effort among research groups in the worldwide modeling community. We believe there is likely to be significant duplication of effort due to the lack of compatibility of models produced by different research groups, and even within research groups. Imagine that each of these models is created five times by the major research groups doing microbiome modeling: the cost has just increased to $25M!

We see four potential strategies for decreasing the costs of microbiome metabolic modeling: (1) Decrease the time required to create models for individual organisms. (2) Minimize the duplica- tion of effort in creating such models (put another way: ensure that as many individual models as possible are reusable by the worldwide COBRA community). (3) Minimize the cost of com- bining models for individual organisms into models of organism communities. Even if models are reusable in some sense, this integration process may not be trivial. (4) Minimize the cost of using metabolic models by providing powerful and user friendly modeling software tools. Build- ing models also entails running and testing them, and ultimately the time and effort required to utilize microbiome models will depend on these usability factors.

The preceding discussion raises the question of just how long does it really take to build a genome- scale metabolic model? We expect it will vary based on a number of factors including the type of organism (e.g., prokaryotic versus eukaryotic; familiar versus exotic metabolism), the modeling software package being used, the expertise of the person performing the modeling, and the degree of validation of the model. We call for more studies of this issue because it is likely that a better understanding of what factors influence model-building time, and of where the most time is spent during model building, will allow us to decrease the time required for model building.

We next review aspects of SRI’s Pathway Tools software, and in particular of its MetaFlux model- ing component, that address the preceding four strategies for minimizing the costs of microbiome modeling. We posit that Pathway Tools is one of the most comprehensive and highly engineered software tools for metabolic reconstruction and modeling.

Extended Abstract: File Not Uploaded

See more of this Session: Multiscale Modelling
See more of this Group/Topical: General Submissions