A novel machine learning model based on ubiquitin-related gene pairs and clinical features to predict prognosis and treatment effect in colon adenocarcinoma

Abstract

Background: Ubiquitin and ubiquitin-like (UB/UBL) conjugations are essential post-translational modifications that contribute to cancer onset and advancement. In colon adenocarcinoma (COAD), nonetheless, the biological role, as well as the clinical value of ubiquitin-related genes (URGs), is unclear. The current study sought to design and verify a ubiquitin-related gene pairs (URGPs)-related prognostic signature for predicting COAD prognoses.

Methods: Using univariate, least absolute shrinkage and selection operator (LASSO), and multivariate Cox regression, URGP's predictive signature was discovered. Signatures differentiated high-risk and low-risk patients. ROC and Kaplan-Meier assessed URGPs' signature. Gene set enrichment analysis (GSEA) examined biological nomogram enrichment. Chemotherapy and tumor immune microenvironment were also studied.

Results: The predictive signature used six URGPs. High-risk patients had a worse prognosis than low-risk patients, according to Kaplan-Meier. After adjusting for other clinical characteristics, the URGPs signature could reliably predict COAD patients. In the low-risk group, we found higher amounts of invading CD4 memory-activated T cells, follicular helper T cells, macrophages, and resting dendritic cells. Moreover, low-risk group had higher immune checkpoint-related gene expression and chemosensitivity.

Conclusion: Our research developed a nomogram and a URGPs prognostic signature to predict COAD prognosis, which may aid in patient risk stratification and offer an effective evaluation method of individualized treatment in clinical settings.

Keywords: Colon adenocarcinoma; Prognostic signature; Tumor immune microenvironment.

Fig. 1

Identification and establishment of the URGPs signature in COAD. A Soft-thresholding power in WGCNA. B Tree of gene clusters. The dynamic tree cutting approach was applied to discover modules by separating the tree diagram at significant branch points. This was premised on an adjacency-based mismatch that was found in the hierarchical gene clustering chart. In the horizontal bar immediately below the tree diagram, various colors have been designated for each module. C Associations between modules and traits in normal and malignant tissues. The table is organized such that each row signifies a color module while each column signifies a clinical characteristic. The correlation coefficient between each module and clinical features and the p-value corresponding to that coefficient is shown by the numbers in each cell. D The forest plot depicting the prognostic-associated URGPs as determined by the univariate Cox proportional hazards regression model in COAD patients. E The calculation of penalties by one thousand rounds of cross-validation to get the optimal values for the parameters. F LASSO-Cox regression analysis was performed by computing the minimal criterion

Fig. 2

The expression of the genes involved in COAD patients' prognoses. A The expression profiles of the 3 genes in COAD and normal samples of the colon. Wilcoxon rank-sum tests were carried out to analyze the differences in the levels of gene expression that were observed between the tumor and the normal samples. ***p < 0.001. B A heat map of gene expression in the low- and high- risk groups. *p < 0.05, **p < 0.01, ***p < 0.001. C Immunohistochemistry images of 2 URGs (OTUB2, RASD2) in COAD and normal samples of the colon. D Sankey diagrams representing the potential regulatory relationships of URGs and TFs. E Sankey diagrams representing the potential regulatory relationships of URGs and eRNAs

Fig. 3

Assessment and confirmation of the predictive significance of the URGPs signature in COAD. A Plots representing the Kaplan–Meier overall survival data for the training group depending on the risk scores. B Plots of overall survival calculated using Kaplan–Meier for the test group as per risk scores. C The Kaplan–Meier plots show the overall survival rate in relation to the risk scores for the whole group. D In the training set, the ROC for overall survival was calculated. E ROC representing the overall survival rate of the test group. F ROC measures survival rates in the whole group. G The risk score distribution in the training group. H The risk score distribution in the test group. I The risk score distribution in the whole group. J Plot depicting the survival rates of patients belonging to the training group. K Survival plots of patients in the test group. L Plots showing the survival of patients in the whole group

Fig. 4

A nomogram that incorporates both clinical and pathological variables and the URGPs signature. A Univariate Cox regression examination of OS-related factors. B OS-related factors subjected to a multivariate Cox regression analysis. C Wilcoxon rank-sum test showed COAD risk scores were associated to clinical stage. D Wilcoxon rank-sum test showed COAD risk scores were connected to lymph node metastasis. E The nomogram for prognostic prediction in COAD. F The nomogram-based ROC curve analysis displays 1-, 3-, and 5-year OS and the corresponding AUC values for COAD patients from the TCGA cohort. G The calibration curve used to verify the predictive performance of the model

Fig. 5

URGPs signature association with biological functions. A, B KEGG findings for the low- and high-risk groups. C, D Findings of the GOBP for the low- and high-risk groups. E, F GSEA reveals the hallmark pathways enriched in the low- and high-risk groups of the URGPs signatures in COAD

Fig. 6

Correlation of URGPs risk score with MSI, TMB, and drug sensitivity in COAD patients. A The immunity infiltration difference between high- and low-risk scores. B The expression of immune checkpoint genes in the high- and low-risk groups. C Box plot depicting ferroptosis-related gene expression in two groups. D Box plot illustrating the expression of pyroptosis-related genes in two groups. E, F Relationship between the URGPs risk score and the MSI. G Comparison of TMB differences in the high- and low-risk groups. H Pearson correlation analysis of URGPs risk score and TMB. I–M The IC₅₀ of 5 routinely used chemotherapy-related drugs (cisplatin, cytarabine, dasatinib, docetaxel and gemcitabine)