Machine Learning Framework Reveals Gastric Microbes Associated with Gastric Inflammation

Previous evidence have associated microbial alterations with gastritis, but questions about the replicability of biomarkers across cohorts and populations remain poorly resolved. Patients from different regions provides most informative, which is conducive to exploration of gastric microbes associated with gastric inflammation. We performed a meta-analysis of 12 available datasets, considering in total 367 fecal 16s rRNA after the quality control and discarding samples of gastric fluid. During exploring atrophic gastritis (AG) signatures, we discarded samples of healthy control since the gastric microbiome in healthy control showed difference than superficial gastritis (SG). The microbial diversity of AG was remarkably distinguished from SG, and the microbial composition of SG tended to AG. According to investigation of validation and discovery cohorts, Bacteroides, Weissella, Actinomyces, Atopobium, Oribacterium, Peptostreptococcus, and Rothia were biomarker between SG and AG without H. pylori infection while Actinomyces, Cutibacterium, Helicobacter, Novosphingobium, Pseudomonas, and Streptococcus are signatures between SG and AG with H. pylori infection. We also found that the development of AG_N is associated with lower vitamin B12 while Significant changes in the gastric microbiota functions of AG_P is gallate and amino acid degradation and heme biosynthesis. Also, non-dialysis classifiers were came up for AG_N with area under the curve (AUC) of 0.7417 and AG_P with AUC of 0.8862. In conclusion, this study comprehensively characterizes gut microbiome in non-dialysis AG and SG and demonstrates the potential of microbial markers as non-invasive diagnostic tools for AG worldwide.