Multi-Omics Integration: Immune-Active Colorectal Cancer

Immune checkpoint blockade rescues only a sliver of colorectal cancer patients, the roughly 15 percent whose tumors are microsatellite instability-high or mismatch-repair-deficient. The microsatellite-stable majority mostly resists. A study in Molecular Therapy: Nucleic Acids applied multi-omics integration to the TCGA colorectal adenocarcinoma cohort to test whether that binary overlooks immune-hot tumors hiding inside the stable group. Pulling together messenger RNA, long noncoding RNA, DNA methylation, and somatic mutation profiles from 275 patients, the team resolved three molecular subtypes and flagged one, called CS3, as both genomically unstable and immunologically inflamed.

Why Multi-Omics Integration Beats Single-Layer Subtyping

Consensus clustering through the MOVICS framework, which pools ten algorithms into one call, settled on three subtypes whose separation held up across silhouette width, the gap statistic, and the consensus prediction index. CS1 leaned metabolic. CS2 was the proliferative one, busy with RNA processing, while CS3 lit up with cytokine signaling and interferon response. Overall survival differed across the groups (log-rank p = 0.002), and the assignments reproduced in the independent GSE39582 cohort with Kappa above 0.6 by nearest template prediction. Older transcriptome-only schemes capture part of this picture, but they miss the cross-omics layering that lets methylation and mutation data explain immune states the RNA cannot. Building on multi-omics clustering tools like MOVICS (Lu et al., 2021), the authors treated each layer as complementary rather than redundant.

Key Findings

• Three subtypes, one clear immune outlier: Integrative clustering of mRNA, lncRNA, methylation, and mutations split 275 colorectal tumors into CS1, CS2, and CS3, with CS3 carrying the strongest cytokine and interferon signaling and the heaviest infiltration of T cells, NK cells, macrophages, and dendritic cells.

• CS3 wears the marks of genomic instability: These tumors showed elevated tumor mutation burden, neoantigen burden, and MSI scores alongside lower intratumoral heterogeneity, with frequent mutations in large structural genes such as MUC16 (77 percent), TTN, and SYNE1 rather than the canonical APC and KRAS drivers.

• Some microsatellite-stable tumors are secretly immunogenic: Within the clinically microsatellite-stable population, a small CS3 subgroup turned out to be ultra-mutated, over 2,000 mutations in three of four cases, the kind of immune-hot tumor that routine MSI testing would pass over.

• An 11-gene signature for risk: Benchmarking 118 survival models picked a random survival forest plus plsRcox hybrid, yielding an 11-gene signature that split patients into high- and low-risk groups; high-risk tumors showed predicted sensitivity to dasatinib, trametinib, and the HDAC inhibitor romidepsin.

Figure 1. Identification and validation of three colorectal cancer molecular subtypes built from multi-omics integration. Panel A is a heatmap of subtype-specific mRNA, long noncoding RNA, DNA methylation, and somatic mutation features. Panel B shows agreement across ten clustering algorithms. Panel C is the consensus matrix defining CS1 through CS3. Panel D gives Kaplan-Meier overall survival curves separating the subtypes (log-rank p = 0.002). Panel E validates the assignments in the external GSE39582 cohort by nearest template prediction. Adapted from Zou et al. (2026), Molecular Therapy: Nucleic Acids.

Genomic Instability Feeding an Inflamed Microenvironment

The genomic picture under CS3 was disordered in a useful way. High mutation and neoantigen load came with raised expression of checkpoint genes including CD274 (PD-L1) and CTLA4, plus broader HLA expression that points to stronger antigen presentation. The T cell inflamed signature was much higher in CS3 (p < 0.0001), and subclass mapping tied these tumors to melanoma cohorts that responded to anti-PD-1 and anti-CTLA-4 therapy. Single-cell RNA-seq across 164,355 cells showed the CS3 program spilling beyond epithelial cells into T/NK and myeloid compartments, where macrophage subsets mixed pro-inflammatory and immunosuppressive wiring at once. That paradox, inflammation running next to immune evasion, is probably why some of these tumors still slip past treatment.

An 11-Gene Signature for Risk and Drug Sensitivity

Turning a research subtype into something a clinic could use meant shrinking the input. RNA sequencing alone recovered most of the multi-omics labels (Kappa above 0.6), a practical floor for assigning CS3 without running every assay. From the CS3 gene set the team benchmarked 118 prognostic models and settled on a random survival forest plus plsRcox hybrid, distilled to 11 hub genes that held prognostic weight across three cohorts. High-risk scores tracked with proliferative and epithelial-mesenchymal transition programs and with an immune-excluded microenvironment. Pharmacogenomic screening across GDSC, CTRP, and PRISM flagged candidate vulnerabilities for high-risk tumors, among them SRC and multi-kinase inhibitors, MEK inhibitors, and epigenetic agents.

From Many Omics Layers to One Decision

What sells this analysis is that no single layer carried the story; methylation and mutation data explained immune states the transcriptome alone could not reach. Generating those layers from the same sample, the way the Omni-MS workflow at Dalton is designed to do, avoids the batch effects and split-aliquot tradeoffs that creep in when each omics layer comes off a different platform on a different day. For a stratification signature meant to survive the trip into the clinic, that internal consistency often matters more than the next clever algorithm.

Frequently Asked Questions

What is multi-omics integration?

Multi-omics integration combines several molecular data layers, such as transcriptomics, DNA methylation, and mutation profiles, from the same samples into one analysis. It captures biology any single layer would miss, which in cancer supports more stable patient subtyping than single-omics methods.

How does multi-omics integration improve colorectal cancer immunotherapy selection?

By layering mutation, methylation, and expression data, the CS3 subtype caught immune-hot, hypermutated tumors inside the microsatellite-stable group that standard MSI testing would call non-responders. Multi-omics integration here refines, rather than replaces, current MSI and consensus molecular subtype calls.

What samples and data are needed for a multi-omics colorectal cancer study?

This work reused public TCGA and GEO data spanning bulk RNA, long noncoding RNA, methylation, somatic mutations, and single-cell RNA-seq, so no new sequencing was required. The authors also showed RNA sequencing alone can approximate the full labels, which lowers cost for routine use.

Conclusion

CS3 reads like a real refinement of how immune-hot colorectal cancer gets defined, and the case for screening microsatellite-stable tumors for hidden hypermutation is convincing. The 11-gene signature and drug predictions, drawn from public cohorts and cell-line pharmacology, still need prospective and immune-competent validation before they guide care. The near-term value is practical: an RNA-based readout that can pull a handful of overlooked patients toward immunotherapy trials.

Citation

Zou, M., Wang, Z., Zhou, W., Xue, G., Hui, Y., Pang, F., Tan, R., Xu, Z., Jin, X., Sun, H., Wang, P., & Jiang, Q. (2026). Multi-layer molecular profiling defines an immune-active colorectal cancer subtype with therapeutic relevance. Molecular Therapy. Nucleic Acids, 37(2), 102907. https://doi.org/10.1016/j.omtn.2026.102907

Note

This blog post summarizes findings from the above-cited research. Figures are adapted from the original publication. For full details, please refer to the source article.

By Seungjun Yeo, CEO at Dalton Bioanalytics. Specializing in multi-omics mass spectrometry for drug discovery and biomarker research.