Metagenomic analysis of common intestinal diseases reveals relationships among microbial signatures and powers multi-disease diagnostic models

Abstract

Intestinal diseases can share clinical symptoms and gut microbiome alterations, necessitating cross-disease comparisons and the use of multi-disease models. Here, we performed meta-analyses on fecal metagenomes of Crohn's disease (CD), ulcerative colitis (UC) and colorectal cancer (CRC), and identified 65 marker species and 74 marker pathways that were significantly altered in >=2 studies of the same disease. We grouped marker species by their co-occurrences into disease-specific and mixed clusters according to whether all the cluster members belong to the same disease, and revealed their distinct phylogenetic relationships: species in disease-specific clusters are phylogenetically related, while those in mixed clusters are more distant; strikingly, UC- and CRC-specific species are phylogenetically closer than to CD-specific ones, likely because UC-patients have higher risk of CRC. Consistent to their phylogenetic relationships, marker species had similar within-cluster and different between-cluster metabolic preferences. In CD and CRC, most marker species correlated with an indicator of leaky gut, suggesting a link between gut dysbiosis and human derived contents. In fact, our results suggested that due to physio-metabolic alterations, the diseased gut may represent a stressed environment and pose stronger selection to gut microbes; consequently, marker species in diseases showed more coordinated changes than the controls. With the marker species and pathways, we constructed three high-performance (including multi-disease) models with AUROC of 0.88 and true positive rates up to 89%, and explained their putative clinical applications. Together, our results improved our understanding on the relationships among common intestinal diseases, and strongly support the feasibility of metagenome-based multi-disease diagnosis.