Learning machine approach reveals microbial signatures of diet and sex in dogOriginal paper
Karen Pendergrass is a microbiome researcher specializing in microbiome-targeted interventions (MBTIs). She systematically analyzes scientific literature to identify microbial patterns, develop hypotheses, and validate interventions. As the founder of the Microbiome Signatures Database, she bridges microbiome research with clinical practice. In 2012, based on her own investigative research, she became the first documented case of FMT for Celiac Disease, four years before the first published case study.
Read MoreWhat was studied?
This study tested whether diet and sex leave detectable signatures in the fecal microbiome of healthy dogs. It pooled data from eight dietary intervention studies run under one standardized protocol. Stool bacteria were profiled by 16S ribosomal RNA gene sequencing of the V3 and V4 regions. Relative abundances of genera were fed into linear discriminant analysis and then random forest classification. A Kruskal-Wallis test confirmed genus-level differences.
Who was studied?
The dataset held 340 fecal samples from 132 healthy adult dogs. Samples were collected serially during diet trials in kennels, shelters, and private homes. Diets fell into four types: commercial extruded, commercial moist, home-made, and a raw meat base diet. Dogs were grouped by sex as intact males, castrated males, intact females, and spayed females. This was an animal microbiome study, not a human one.
What were the most important findings?
Firmicutes was the most abundant phylum overall, averaging 713.9 per thousand, and was highest on the home-made diet. Bacteroidetes averaged 103.4 and Fusobacteria 57.5 per thousand, with Fusobacteria highest on the moist diet. Random forest classified dogs by diet with 75.29% overall accuracy. Home-made and base diets overlapped and were merged, while commercial moist reached 88.24% correct classification. Sex also structured the microbiome. Castrated males and spayed females clustered together and separated from intact dogs, which PERMANOVA confirmed at p below 0.05.
What are the greatest implications of this study?
The results show diet and sex both shape canine gut bacteria strongly enough for machine learning to sort dogs by these factors. This supports using microbial profiles as fingerprints of feeding and reproductive status. The clustering of neutered animals hints at a two-way link between gut bacteria and host hormones. Limitations include underrepresentation of the home-made diet and of intact males and females. The authors call these preliminary results needing larger, multi-lab datasets.