An Ecology of Modeling Methods, Data, and Tools to Decipher Functional Delegation and Species Interactions within a Microbiome

Authors: 
Henry, C. S., Argonne National Laboratory
Weisenhorn, P., Argonne National Laboratory
van der Lelie, D., Gusto Global
Jeffryes, J. G., Argonne National Laboratory
Chivian, D. C., Joint BioEnergy Institute
Faria, J. P., Argonne National Laboratory
Edirisinghe, J. N., Argonne National Laboratory
Liu, F., Argonne National Laboratory
Seaver, S. M. D., Argonne National Laboratory
Dudley, H., University of Colorado Boulder
Zhang, Q., Argonne National Laboratory
Sadkhin, B., Argonne National Laboratory
Gupta, N., Argonne National Laboratory
Gu, T., Argonne National Laboratory
Taylor, R. C., Pacific Northwest National Laboratory
Song, H. S., Pacific Northwest National Laboratory
Bernstein, H. C., Pacific Northwest National Laboratory
Zucker, J., Pacific Northwest National Laboratory
Lindemann, S. R., Pacific Northwest National Laboratory
Gilbert, J., Argonne National Laboratory
Cottingham, R., Oak Ridge National Laboratory
Arkin, A., University of California, Berkeley
The term metagenome was first used in the literature 20 years ago, and in that time, we have witnessed a revolution in our capacity to gather data, model, and understand microbiome systems in natural and engineered environments. Finally we have sufficient data available to support the validation and refinement of metabolic models for microbiome systems. Here, I describe an ecology of tools, pipelines, and data that we have developed to support the modeling and simulation of microbiome systems. One of the initial challenges in modeling a microbiome system is to obtain species genomes from metagenomic data. To aid with this challenge, the KBase team recently implemented a pipeline for metagenome assembly, binning, QC, and annotation comprised of available open-source tools (e.g. metaSPAdes, MaxBin2). Another challenge is to rapidly construct predictive genome-scale models from newly assembled and binned genomes. To address this challenge, we have developed publicly accessible tools to generate interoperable models for microbial, plant, and fungal genomes (with recent significant improvements to our microbial model reconstruction pipeline). Even with models, these community systems are typically under-determined, with species interactions difficult to identify from noise. For this, we developed tools to predict auxotrophy from genomic data, revealing exciting new insights into how microbiome systems evolve and build stable interconnections. We also developed tools and pipelines to support the mapping of models and microbiome systems in general to metabolomics data, applying network approaches to enable models to slice through noise. Finally, we recently developed a new approach that explores the balance of protein-resources that plays a significant role in the delegation of functions in microbiome systems. In my talk, I describe these tools and explore the results from their application to a variety of microbiome systems, including a photo-autotrophic microbial mat, an electrosynthetic microbiome, and a microbiome on a built surface.