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

Abstract

AbstractMicrobial metabolic processes greatly impact ecosystem functioning and the physiology of multi-cellular host organisms. The inference of metabolic capabilities and phenotypes from genome sequences with the help of reference biomolecular knowledge stored in online databases remains a major challenge in systems biology. Here, we present gapseq: a novel tool for automated pathway prediction and metabolic network reconstruction from microbial genome sequences. gapseq combines databases of reference protein sequences (UniProt, TCDB), in tandem with pathway and reaction databases (MetaCyc, KEGG, ModelSEED). This enables the prediction of an organism’s metabolic capabilities from sequence homology and pathway topology criteria. By incorporating a novel LP-based gap-filling algorithm, gapseq facilitates the construction of genome-scale metabolic models that are suitable for metabolic phenotype predictions by using constraint-based flux analysis. We validated gapseq by comparing predictions to experimental data for more than 3, 000 bacterial organisms comprising 14, 895 phenotypic traits that include enzyme activity, energy sources, fermentation products, and gene essentiality. This large-scale phenotypic trait prediction test showed, that gapseq yields an overall accuracy of 81% and thereby outperforms other commonly used reconstruction tools. Furthermore, we illustrate the application of gapseq-reconstructed models to simulate biochemical interactions between microorganisms in multi-species communities. Altogether, gapseq is a new method that improves the predictive potential of automated metabolic network reconstructions and further increases their applicability in biotechnological, ecological, and medical research. gapseq is available at https://github.com/jotech/gapseq.