Cross-cohort resistome and virulome gradients structure the colorectal cancer microbiome.

The gut microbiome is increasingly implicated in colorectal cancer (CRC), yet the functional signatures associated with disease progression remain poorly resolved across populations. We performed an assembly-based metagenomic analysis of more than 500 samples from three geographically distinct cohorts to characterize resistome and virulome patterns associated with CRC. Using a cross-validated modeling framework based on Partial Least Squares (PLS) regression, we identified two reproducible latent functional gradients that structured variation in antimicrobial-resistance and virulence-factor profiles. One gradient was enriched for adhesion, efflux, and biofilm-associated functions, while the second reflected immunomodulatory and barrier-related pathways. These components were statistically robust, directionally stable across cohorts, and consistent with functional themes frequently reported in CRC microbiome studies. To summarize variation along these gradients, we derived an exploratory Dual-Axis Index (DAI) based on the two stable PLS components. Although its discriminative performance was moderate, the DAI provided an interpretable low-dimensional representation of how resistome-virulome patterns differed across healthy, adenoma, and carcinoma states. These results suggest that functional gene profiles in CRC are organized along reproducible statistical axes, and highlight functional modules, such as adhesion-, iron-associated, and immune-interaction pathways that may complement taxonomic or metabolic biomarkers in future multimodal approaches. Our work provides a reproducible, assembly-based framework for examining the functional organization of CRC-associated microbiomes across diverse populations.