Multimodal meta-analysis of fecal metagenomes reveals microbial single nucleotide variants as superior biomarkers for early detection of colorectal cancer

Microbial signatures show remarkable potentials in predicting colorectal cancer (CRC). This study aimed to evaluate the diagnostic powers of multimodal microbial signatures, multi-kingdom species, genes, and single-nucleotide variants (SNVs) for detecting pre-cancerous adenomas. We performed cross-cohort analyses on whole metagenome sequencing data of 750 samples via xMarkerFinder to identify adenoma-associated microbial multimodal signatures. Our data revealed that fungal species outperformed species from other kingdoms with an area under the ROC curve (AUC) of 0.71 in distinguishing adenomas from controls, while classifier based on SNVs of bacterial species displayed the strongest diagnostic capability (AUC=0.89). SNV biomarkers also exhibited outstanding performances in three independent validation cohorts (AUCs = 0.83, 0.82, and 0.76, respectively) and were highly specific to adenoma. In further support of the above results, functional analyses revealed more frequent inter-kingdom associations between bacteria and fungi, and abnormalities in quorum sensing, purine and butanoate metabolism in adenoma, which were further validated in a newly-recruited cohort via qRT-PCR. Therefore, these data extend our understanding of adenoma-associated multimodal alterations in the gut microbiome and provide a rationale of microbial SNVs for the early detection of CRC.