Transcriptome-guided parsimonious flux analysis improves predictions with metabolic networks in complex environments

Abstract

The metabolic responses of bacteria to dynamic extracellular conditions drives not only the behavior of single species, but also entire communities of microbes. Over the last decade, genome-scale metabolic network reconstructions have assisted in our appreciation of important metabolic determinants of bacterial physiology. These network models have been a powerful force in understanding the metabolic capacity that species may utilize in order to succeed in an environment. Increasingly, an understanding of context-specific metabolism is critical for elucidating metabolic drivers of larger phenotypes and disease. However, previous approaches to use network models in concert with omics data to better characterize experimental systems have met challenges due to assumptions necessary by the various integration platforms or due to large input data requirements. With these challenges in mind, we developed RIPTiDe (Reaction Inclusion by Parsimony and Transcript Distribution) which uses both transcriptomic abundances and parsimony of overall flux to identify the most costeffective usage of metabolism that also best reflects the cell's investments into transcription. Additionally, in biological samples where it is difficult to quantify specific growth conditions, it becomes critical to develop methods that require lower amounts of user intervention in order to generate accurate metabolic predictions. Utilizing a metabolic network reconstruction for the model organism Escherichia coli str. K-12 substr. MG1655 (iJO1366), we found that RIPTiDe correctly identifies context-specific metabolic pathway activity without supervision or knowledge of specific media conditions. We also assessed the application of RIPTiDe to in vivo metatranscriptomic data where E. coli was present at high abundances, and found that our approach also effectively predicts metabolic behaviors of host-associated bacteria. In the setting of human health, understanding metabolic changes within bacteria in environments where growth substrate availability is difficult to quantify can have large downstream impacts on our ability to elucidate molecular drivers of disease-associated dysbiosis across the microbiota. Our results indicate that RIPTiDe may have potential to provide understanding of context-specific metabolism of bacteria within complex communities.

An updated genome-scale metabolic network reconstruction ofPseudomonas aeruginosaPA14 to characterize mucin-driven shifts in bacterial metabolism

“…The main difference between the strains is the presence of additional gene clusters in PA14 (most linked to virulence) that we would not expect to have a large effect on overall metabolism. PAO1 genes in the transcriptomic dataset were mapped to PA14 orthologs and then the data was integrated with the iPau21 using the RIPTiDe algorithm 36 . RIPTiDe uses transcriptomic evidence to create context-specific metabolic models representative of a parsimonious metabolism consistent with the transcriptional investments of an organism.…”

“…Published transcriptomic data was integrated with the model using RIPTiDe 36 . The transcriptomic data was normalized then translated from PAO1 genes to the orthologous PA14 genes prior to integration 59 .…”

Mechanistic models of microbial community metabolism

“…Several algorithms have been developed to integrate transcriptomic and proteomic data with metabolic models. 17–20 These methods typically convert gene expression or protein abundance levels to reaction weights that represent the likelihood that individual reactions are being utilized by the organism. By applying these weights to the model, the predicted metabolic state can be guided towards what is observed experimentally.…”

“… 54 RIPTiDe and TIMBR are both integration algorithms that utilize transcriptomic data to create reaction weights, which then impact predicted model metabolic outputs. 17,55 RIPTiDe additionally allows for this transcriptomic data to be used to “prune” reactions and metabolites from a model that do not have strong transcriptional support and are not necessary for an OF to carry flux. However, due to the varying residence time of RNA (∼3 minutes) and proteins (0.5–35 hours), these data offer different temporal snapshots of what is occurring in the organisms.…”

Novel drivers of virulence inClostridioides difficileidentified via context-specific metabolic network analysis

“…As previously stated, GENREs have provided powerful platforms for the integration of transcriptomic data, creating greater context for the shifts observed between conditions and capturing the potential influence of pathways not obviously connected 81 . With this application in mind, we chose to generate context-specific models for both in vitro and in vivo experimental conditions characterized with RNA-Seq analysis utilizing a recently published unsupervised transcriptomic data integration method 82 . Briefly, this approach calculates the most cost-efficient usage of the metabolic network in order to achieve growth given the pathway investments indicated by the transcriptomic data.…”