Background
Colorectal cancer (CRC) is a leading global cause of cancer-related mortality, and gut microbiome dysbiosis has been increasingly implicated in its development. Prior studies have predominantly focused on bacterial communities, while the contributions of other microbial kingdoms — including fungi, archaea, and viruses — and their interactions across diverse geographical populations remain underexplored.
Objectives
This study aimed to identify reproducible multi-kingdom microbial signatures associated with CRC across geographically distinct cohorts, to assess their diagnostic potential, and to characterise ecological interactions among microbial kingdoms that may contribute to disease pathogenesis.
Methods
We integrated shotgun metagenomic sequencing data from 8 independent cohorts, totalling 1,368 stool samples from individuals with CRC and tumour-free controls. A uniform analytical pipeline was applied to reduce study heterogeneity. Differential abundance analyses were conducted for archaeal, bacterial, fungal, and viral species, and machine-learning diagnostic models were developed using single-kingdom and combined multi-kingdom features. Ecological co-abundance networks were analysed to identify cross-kingdom interactions, and functional profiling was undertaken to explore metabolic potentials associated with CRC.
Results
Distinct microbiota alterations were observed across all four kingdoms in CRC patients compared with controls. Individual single-kingdom diagnostic models achieved moderate performance, with bacterial signatures showing the highest area under the receiver operating characteristic curve (AUROC). Combined multi-kingdom models substantially improved diagnostic accuracy, with a minimal panel of 16 markers (11 bacterial, 4 fungal, 1 archaeal species) achieving AUROC ≈ 0.83 across three independent validation cohorts. Co-abundance network analysis revealed significant bacterial–fungal associations, such as between Talaromyces islandicus and Clostridium saccharobutylicum, suggesting potential inter-kingdom ecological links in CRC. Functional metagenomic profiling highlighted elevated metabolic pathways, including D-amino acid and butanoate metabolism, in CRC-associated microbiomes.
Conclusions
This multi-cohort metagenomic analysis reveals reproducible and diagnostically relevant multi-kingdom microbiota signatures of CRC that transcend bacterial communities alone. Integration of bacterial, fungal, archaeal, and viral features enhances diagnostic performance and underscores the importance of inter-kingdom microbial interactions in disease. These findings support the development of non-invasive, multi-kingdom biomarker panels for CRC detection and prompt further investigation into the mechanistic roles of microbiome diversity in colorectal carcinogenesis.