Gempipe: a tool for drafting, curating, and analyzing pan and multi-strain genome-scale metabolic models

Genome-scale metabolic models (GSMMs) can mechanistically explain phenotypic differences among closely related bacterial strains. However, high-throughput multi-strain reconstructions of GSMMs are still challenging: reference-based methods inherit curated information while missing new contents; alternatively (universe-based), reference-free methods could cover strain-specific reactions, but they disregard curated information. Ideally, references should be curated pan-GSMMs for species (or genus), but their reconstruction is extremely demanding, making them still rare in the literature. Here, Gempipe is presented, a computational tool streamlining the multi-strain reconstruction and analysis of GSMMs, going through the production of a pan-GSMM. Its reconstruction method is hybrid; as an optional reference, GSMM is automatically expanded with extra reactions taken from a reference-free reconstruction. Gempipe also downloads, filters, and annotates genomes; performs in-depth gene recovery; annotates models' contents; and predicts strain-specific capabilities. The companion programming interface includes functions ranging from the (pan-)GSMMs' curation to the multi-strain analysis. Gempipe was validated using multi-strain data sets, showing improved accuracy when compared with state-of-the-art tools. Moreover, metabolic diversities within Limosilactobacillus reuteri were explored, grouping strains into metabolically coherent clusters and systematically predicting health-related metabolites' biosynthesis.IMPORTANCEAvailable genome-scale metabolic model (GSMM) reconstruction tools present major limitations in the context of multi-strain modeling. Gempipe surpasses these limitations by implementing a novel, hybrid reconstruction strategy. Not only does it produce more accurate strain-specific GSMMs, but it also produces pan-GSMMs when the only available reference is a manually curated model for a single strain, which is currently the most common case. With the vast availability of genome sequences, the high-throughput, multi-strain GSMM reconstruction and analysis approach provided by Gempipe will facilitate large-scale studies of exploration and bioprospecting of strain-level bacterial metabolic diversity, moving a step forward in strains' screening and rational selection.