Fast automated reconstruction of genome-scale metabolic models for microbial species and communities

Abstract

Genome-scale metabolic models are instrumental in uncovering operating principles of cellular metabolism, for model-guided re-engineering, and unraveling cross-feeding in microbial communities. Yet, the application of genome-scale models, especially to microbial communities, is lagging behind the availability of sequenced genomes. This is largely due to the time-consuming steps of manual curation required to obtain good quality models. Here, we present an automated tool, CarveMe, for reconstruction of species and community level metabolic models. We introduce the concept of a universal model, which is manually curated and simulation ready. Starting with this universal model and annotated genome sequences, CarveMe uses a top-down approach to build single-species and community models in a fast and scalable manner. We show that CarveMe models perform closely to manually curated models in reproducing experimental phenotypes (substrate utilization and gene essentiality). Additionally, we build a collection of 74 models for human gut bacteria and test their ability to reproduce growth on a set of experimentally defined media. Finally, we create a database of 5587 bacterial models and demonstrate its potential for fast generation of microbial community models. Overall, CarveMe provides an open-source and user-friendly tool towards broadening the use of metabolic modeling in studying microbial species and communities.

Metage2Metabo: metabolic complementarity applied to genomes of large-scale microbiotas for the identification of keystone species

“…GSMN of interest after M2M analysis can be further analysed and refined if needed. Nevertheless, there are several other software available for GSMN reconstruction in addition to Pathway Tools such as Kbase [3], Mod-elSEED [21] or CarveMe [32]. Metabolic networks resulting from such platforms can be used as inputs to M2M for metabolic analysis as the reconstruction step can be bypassed and the tool accepts GSMNs in SBML format [23], widely used for exchanging and distributing models.…”

gapseq: Informed prediction of bacterial metabolic pathways and reconstruction of accurate metabolic models

“…table S1). All reactions and metabolites from the database were included for the construction of a full universal metabolic network model; an approach that is also used in CarveMe [50]. We curated the underlying biochemistry database in order to correct inconsistencies in reaction stoichiometries and reversibilities.…”

“…This approach is especially relevant in cases where the sequence similarity to known enzyme-coding genes was close to but did not reach the cutoff value b, which is required for a reaction to be included directly into the draft network. In contrast to the gap-filling algorithms described in previous works [4] and [50], which also use genetic evidence-weighted gap-filling, the gap-filling problem in gapseq is not formulated as Mixed Integer Linear Program (MILP) but as Linear Program (LP), and is derived from the parsimonious enzyme usage Flux Balance Analysis (pFBA) algorithm developed by Lewis et al, 2010 [47]. Therefore, the alignment statistics (i.e.…”

Expanding Metabolic Capabilities Using Novel Pathway Designs: Computational Tools and Case Studies

“…Once the metabolic network is assembled, metabolic substructures like pathways and subnetworks can be explored using different mathematical methods for a variety of modeling purposes (Figure ). For example, context‐specific networks or subnetworks can be constructed in a top‐down manner, using the supernetwork, as an annotated, high‐quality scaffold for describing the metabolism of an organism or microbial community . By integrating experimental and annotated sequencing data from the specific organism with the network, optimization programs can be formulated that maximize the network connectivity and agreement of inferred subnetwork with the data.…”

“…For example, context-specific networks or subnetworks can be constructed in a top-down manner, using the supernetwork, as an annotated, high-quality scaffold for describing the metabolism of an organism or microbial community. [41,42] By integrating experimental and annotated sequencing data from the specific organism with the network, optimization programs can be formulated that maximize the network connectivity and agreement of inferred subnetwork with the data. Conversely, the reconstructed subnetwork can be inferred following a bottom-up approach, where the annotated genes of a determined organism are individually mapped based on different criteria onto the supernetwork.…”