FiberGrowth Pipeline: A Framework Toward Predicting Fiber-Specific Growth From Human Gut Bacteroidetes Genomes

Dietary fibers impact gut colonic health, through the production of short-chain fatty acids. A low-fiber diet has been linked to lower bacterial diversity, obesity, type 2 diabetes, and promotion of mucosal pathogens. Glycoside hydrolases (GHs) are important enzymes involved in the bacterial catabolism of fiber into short-chain fatty acids. However, the GH involved in glycan breakdown (adhesion, hydrolysis, and fermentation) are organized in polysaccharide utilization loci (PUL) with complex modularity. Our goal was to explore how the capacity of strains, from the Bacteroidetes phylum, to grow on fiber could be predicted from their genome sequences. We designed an in silico pipeline called FiberGrowth and independently validated it for seven different fibers, on 28 genomes from Bacteroidetes-type strains. To do so, we compared the existing GH annotation tools and built PUL models by using published growth and gene expression data. FiberGrowth's prediction performance in terms of true positive rate (TPR) and false positive rate (FPR) strongly depended on available data and fiber: arabinoxylan (TPR: 0.89 and FPR: 0), inulin (0.95 and 0.33), heparin (0.8 and 0.22) laminarin (0.38 and 0.17), levan (0.3 and 0.06), mucus (0.13 and 0.38), and starch (0.73 and 0.41). Being able to better predict fiber breakdown by bacterial strains would help to understand their impact on human nutrition and health. Assuming further gene expression experiment along with discoveries on structural analysis, we hope computational tools like FiberGrowth will help researchers prioritize and design in vitro experiments.