Machine Learning Models To Characterize the Association of the Gut Microbiota with Osteopenia and Osteoporosis: A Multi-Cohort Study.

Emerging evidence suggests that gut microbiota dysbiosis is associated with bone metabolism disorders, including osteopenia (ON) and osteoporosis (OP). However, multi-cohort integrated and association analyses remain underexplored. We conducted a comprehensive meta-analysis of gut microbiota data from six public cohorts, encompassing 341 samples from normal bone density controls (NC), ON, and OP patients. We found neither osteopenia nor osteoporosis patients exhibited significant differences in gut microbial alpha diversity compared to healthy controls. However, Bray-Curtis distance analysis revealed significant beta-diversity differences among groups. We employed a leave-one-cohort-out approach to develop the classification models to link gut microbiota and disease traits. Our analysis revealed that the models achieved accuracies of 72.5-75.6% in classifying ON across two independent cohorts. Furthermore, for osteoporosis OP classification, the models demonstrated accuracies of 70.1%, 71.2%, 80.1%, and 76.6% across four validation cohorts. Collectively, our study identifies distinct gut microbiota signatures in OP/ON, highlighting the importance of several potential SCFAs-producing bacteria.