부산대학교 도서관

OBJECTIVE:: Identifying the gut microbiota associated with host immunity in the AIDS stage. DESIGN:: We performed a cross-sectional study. METHODS:: We recruited people with HIV (PWH) in the AIDS or non-AIDS stage and evaluated their gut microbiota and metabolites by using 16S ribosomal RNA (rRNA) sequencing and liquid chromatography–mass spectrometry (LC-MS). Machine learning models were used to analyze the correlations between key bacteria and CD4 T cell count, CD4 T cell activation, bacterial translocation, gut metabolites, and KEGG functional pathways. RESULTS:: We recruited 114 PWH in the AIDS stage and 203 PWH in the non-AIDS stage. The α-diversity of gut microbiota was downregulated in the AIDS stage (P < 0.05). Several machine learning models could be used to identify key gut microbiota associated with AIDS, including the logistic regression model with area under the curve (AUC), sensitivity, specificity, and Brier scores of 0.854, 0.813, 0.813, and 0.160, respectively. The decreased key bacteria ASV1 (Bacteroides sp.), ASV8 (Fusobacterium sp.), ASV30 (Roseburia sp.), ASV37 (Bacteroides sp.), and ASV41 (Lactobacillus sp.) in the AIDS stage were positively correlated with the CD4 T cell count, the EndoCAb IgM level, 4-hydroxyphenylpyruvic acid abundance, and the predicted cell growth pathway, and negatively correlated with the CD3CD4CD38HLA-DR T cell count and the sCD14 level. CONCLUSION:: Machine learning has the potential to recognize key gut microbiota related to AIDS. The key five bacteria in the AIDS stage and their metabolites might be related to CD4 T cell reduction and immune activation.