A multi-omics prognostic model integrating disulfidptosis-related molecular features in colorectal cancer.

Background

Disulfidptosis is a recently described cell-death-related process associated with intracellular disulfide stress and may be relevant to tumor progression and immune regulation. However, its prognostic significance and broader biological relevance in colorectal cancer (CRC) remain unclear.

Methods

Transcriptomic and clinical data from TCGA and GEO cohorts were systematically integrated to characterize disulfidptosis-related molecular patterns in CRC. A four-gene prognostic signature was developed in a randomly assigned training cohort and evaluated in the corresponding internal test cohort. Public datasets were further used for gene-level expression analysis, and survival associations were assessed in the GSE39582 cohort. Immune infiltration, tumor mutation burden, microsatellite instability, and estimated drug-sensitivity analyses were performed to explore the biological correlates and therapeutic-response-related patterns associated with the model. The four model genes were further examined using single-cell RNA-seq data and public immunohistochemical images from the Human Protein Atlas, and were preliminarily assessed in 11 paired CRC tissues and matched adjacent normal tissues by RT-qPCR and western blotting.

Results

A four-gene signature consisting of SPINK1, GPRASP1, TIMP1, and KRT6A stratified patients into prognostically distinct risk groups. The high-risk group showed poorer survival, higher tumor mutation burden, higher TIDE scores, and elevated stromal, immune, and ESTIMATE scores, suggesting a more immune-evasive tumor microenvironment. Enrichment analyses indicated that high-risk tumors were associated with extracellular matrix remodeling and PI3K-Akt-related pathways, whereas low-risk tumors were enriched in metabolic processes. Estimated drug-sensitivity analyses indicated agent-specific differences between the two risk groups. The high-risk group showed lower estimated IC50 values for several cytotoxic and targeted agents, including cisplatin, docetaxel, doxorubicin, etoposide, vinorelbine, nilotinib, and gefitinib, whereas the low-risk group showed lower estimated IC50 values for AKT inhibitor VIII, sorafenib, lapatinib, erlotinib, and BIBW2992.

Conclusions

This study identified a four-gene prognostic signature associated with disulfidptosis-related molecular heterogeneity in CRC. The signature was associated with survival differences, distinct tumor microenvironmental features, and divergent biological programs across risk groups, while KRT6A showed the most consistent pattern across public-dataset analyses and tissue-level analyses. These findings support the potential relevance of this subtype-guided signature to prognostic stratification in CRC and identify related genes for further biological investigation and independent validation.