Introduction
Recent studies have shown that imbalanced intestinal(肠道的) microbiota, termed “dysbiosis”(菌群失调), contributes to(会导致) various human diseases. The current development of microbial markers has mostly used binary classifiers. Emerging(最新的) evidence, however, suggests that most health conditions exhibit(表现出) overlapping gut microbiome signatures(特征), thus single-disease diagnostic(诊断的) models are likely to be confounded(混淆) by unrelated diseases and may lead to misclassification.
Although an attempt has been made to develop a multi-class diagnostic model, heterogeneity, technical bias and batch effects involved(被涉及) in the previous work relying on public datasets for analyses would limit accuracy.
Here, we develop the largest single-site dataset to date(至今) covering multiple diseases, adopt a machine learning multi-class model to predict different diseases using species-level faecal(粪便的) microbiome profiling(谱), and validate the findings using public metagenome datasets across different populations.
图1：基于粪便微生物组的机器学习用于多类疾病诊断
a.数据集划分、模型训练和独立验证框架。