Machine Learning-Derived Neddylation Gene Signature for Predicting Prognosis and Immunotherapy Benefits in Colorectal Cancer.

Background: Colorectal cancer (CRC) is a major cause of cancer deaths globally, mainly due to treatment resistance. Neddylation, a key post-translational modification, is linked to tumor growth and immune response, offering potential therapeutic targets, though its role in CRC is not well-explored.

Methods: We examined neddylation-related genes (NRGs) across cell subtypes using CRC scRNA-seq data from the TISCH database. Unsupervised clustering of TCGA and GEO bulk RNA-seq data identified various neddylation patterns. A neddylation-related gene signature (NRGS) was developed using ten machine-learning algorithms and validated externally. The study compared biofunctions, including functional analysis, immune cell infiltration, genomic mutations, enrichment analysis, and responses to immunotherapy and chemotherapy, between high- and low-risk groups defined by the NRGS model.

Results: scRNA-seq analysis showed that the high neddylation score group had more malignant and diverse immune and stromal cells, with activated pathways aiding tumor growth and spread. We identified two neddylation patterns: Cluster A and Cluster B. Cluster B, associated with worse survival, had more immunosuppressive cells and increased tumor progression. We developed a neddylation-related gene signature (NRGS) using ten machine-learning algorithms, which accurately predicted outcomes. Higher risk scores correlated with poorer survival, with AUCs of 0.979, 0.989, and 0.996 for 1-year, 2-year, and 3-year OS in the training cohort. The NRGS was linked to higher recurrence or metastasis, advanced disease stage, and independently predicted OS risk. Patients with high NRGS may resist immunotherapy and standard chemotherapy.

Conclusion: The NRGS could predict outcomes and responses to immunotherapy and chemotherapy in CRC patients, aiding personalized treatment, though further validation is needed.