Commercial next-generation sequencing (NGS) platforms in China commonly apply a lung cancer–derived tumor microenvironment (TME) subtyping signature—originally developed in a European cohort—to classify colorectal cancer (CRC). The diagnostic performance of this cross-tumor, cross-population signature in Chinese CRC patients had not been validated. The study aimed to (1) assess the lung cancer–originated TME subtyping system in a Chinese CRC cohort, (2) screen CRC-specific immune mRNA biomarkers for TME subtyping, and (3) evaluate the clinical utility of IRF1, CD8A and LAG3 for distinguishing the immune-enriched (IE) and immune-desert plus fibrotic (D+F) subtypes.
This retrospective analysis enrolled 87 formalin-fixed, paraffin-embedded (FFPE) CRC specimens with complete NGS and clinicopathological data. Patients were classified under the lung cancer–derived TME subtyping system into IE (n=15) and D+F (n=72) groups.
Thirty-one mRNA transcripts were selected for analysis. These transcripts covered two functional modules: 13 genes related to immune–metabolic homeostasis and 18 genes related to immune checkpoint signaling and immune infiltration. The study used Spearman correlation analysis to evaluate co-expression patterns among candidate genes. Receiver operating characteristic (ROC) curves combined with five-fold cross-validation were applied to compare the discriminatory efficacy of single-gene markers and a combined three-gene panel consisting of IRF1, CD8A and LAG3.
The candidate transcripts were grouped into two functional modules reflecting immune-metabolic homeostasis and immune checkpoint/infiltration processes. Among the evaluated genes, IRF1, CD8A and LAG3 demonstrated particularly strong positive co-expression: IRF1–CD8A r=0.93; IRF1–LAG3 r=0.84; CD8A–LAG3 r=0.73. These correlation coefficients indicate high collinearity among the three transcripts in this cohort.
Nominally higher expression of IRF1, CD8A and LAG3 was observed in the IE subtype relative to the D+F subtype. However, after Benjamini–Hochberg false discovery rate (FDR) correction, intergroup differences did not remain statistically significant; the authors attribute this loss of significance principally to the limited number of IE cases in the sample set.
Single-gene ROC analysis produced the following area under the curve (AUC) values: 0.763 for IRF1, 0.752 for CD8A, and 0.771 for LAG3. LAG3 exhibited the best individual discriminatory capacity among the three tested transcripts.
A combined three-gene panel (IRF1 + CD8A + LAG3) was also evaluated using five-fold cross-validation. The cross-validated AUC for the combined panel was 0.717, which was lower than the single-gene AUC for LAG3. The reduced performance of the multi-gene panel was attributed to severe collinearity among the three transcripts, generating redundant predictive information and diminishing incremental discriminatory benefit.
The study concludes that the lung cancer–originated TME subtyping system cannot be directly extrapolated to Chinese CRC patients without validation or adaptation. Among the candidate transcripts tested, LAG3 emerged as a promising independent transcriptomic marker for distinguishing TME subtypes in this Chinese CRC cohort. Due to strong co-expression among IRF1, CD8A and LAG3, combining these transcripts did not improve discrimination and in fact reduced cross-validated performance compared with single LAG3.
The authors recommend that larger, independent multi-center Chinese CRC cohorts be used to develop population-specific immune transcriptomic biomarkers suitable for standardized clinical NGS TME stratification.
The primary limitation reported is the small number of IE cases (n=15), which reduced statistical power and likely contributed to the loss of significance after FDR correction. The study also notes that the severe collinearity among the three top candidate genes limited the value of multi-gene combination models in this dataset.
Recommendations from the authors include conducting large, multi-center validation studies in Chinese CRC populations to derive and standardize immune transcriptomic signatures tailored to this population before integrating such signatures into routine clinical NGS pipelines.
The research protocol received approval from the Ethics Committee of Chinese People's Liberation Army General Hospital (Ethics No.2025080109). The authors declare no competing interests. All data generated in the study are available from the authors upon reasonable request.