Faecal microbiome-based machine learning for multi-class disease diagnosis

Gut microbiome dysbiosis makes substantial contributions to a broad range of health disorders, but its diagnostic implication for most human diseases is largely unknown. Most health conditions exhibit largely overlapped microbial markers, thus single-disease models are likely to be confounded by signals shared across unrelated diseases and may lead to misclassification. Therefore, whether the fecal microbiome-based non-invasive diagnosis is capable of distinguishing multiple diseases is largely unknown. This large cross-sectional study aimed to describe the features of the fecal microbiome of patients under different health conditions and to construct a multiple diseases cohort for developing fecal microbiome-based multi-class diagnosis tools.