Signature of m5C-Related lncRNA for Prognostic Prediction and Immune Responses in Pancreatic Cancer

Abstract

Background: Pancreatic cancer (PC) has a high mortality and dismal prognosis, predicting to be the second most lethal malignancy. 5-Methylcytosine (m5C) and long noncoding RNAs (lncRNAs) are both crucial in the prognostic outcome and immunotherapeutic effect for PC patients. Therefore, we aimed to create an m5C-related lncRNA signature (m5C-LS) for PC patients' prognosis and treatment.

Methods: Clinicopathological information and RNAseq data were acquired from The Cancer Genome Atlas (TCGA) database. Pearson's correlation analysis was used to extract m5C-related lncRNAs in PC. Univariate, least absolute shrinkage and selection operator (LASSO), and multivariate Cox analyses were adopted to build an m5C-LS. Kaplan-Meier (K-M), principal component analysis (PCA), and nomogram were utilized to assess model accuracy. In addition, we explored the model's possible immunotherapeutic responses and drug sensitivity targets.

Results: Three m5C-related lncRNAs were finally established to construct the risk signature, which has a good and independent predictive ability for PC patients. Based on the m5C-LS, patients were classified into the low- and high-m5C-LS group, with the latter having a worse prognosis. Furthermore, the m5C-LS allowed us to better discriminate the immunotherapeutic responses of PC patients in different subgroups.

Conclusions: Our study constructed an m5C-LS and established a nomogram model that accurately predicted the prognosis of PC patients, as well as provides promising immunotherapeutic strategies in the future.

Figure 1

Selection of m5C-related lncRNAs in PC patients. (a) Sankey diagram for the network of m5C genes and related lncRNAs. (b) Heatmap for relationships between 13 m5C genes and 3 m5C-related lncRNAs. ^∗p < 0.05, ^∗∗p < 0.01, and ^∗∗∗p < 0.001.

Figure 2

Description of m5C-LS. (a) The LASSO analysis of PC. (b) Determine the optimal LASSO settings.

Figure 3

Identification of the prognostic value of m5C-LS in TCGA training set. (a) K-M curves of patients' OS between high- and low-m5C-LS groups. (b) Distribution of the risk score and patients. (c) Dot plot of survival status. (d) Heatmap of 3 m5C-related lncRNAs' expression between two groups.

Figure 4

Verification of m5C-LS in TCGA testing and entire sets. (a, e) K-M curves of patients' OS between high- and low-m5C-LS groups. (b, f) Distribution of the risk score and patients. (c, g) Dot plot of survival status. (d, h) Heatmap of 3 m5C-related lncRNAs' expression between two risk groups.

Figure 5

K-M curves of patients' OS grouped by (a) age, (b) gender, (c) TNM stage, and (d) tumor grade between two groups in TCGA entire set.

Figure 6

PCA comparison between two groups based on (a) entire gene profiles, (b) m5C coding genes, (c) m5C-related lncRNAs, and (d) m5C-LS in TCGA entire set.

Figure 7

Evaluation of the m5C-LS prognostic value in TCGA entire set. (a, b) Univariate and multivariate analyses of PC patients' OS. (c) ROC curves of 1-, 3-, and 5-year survival. (d) ROC curves of both clinical features and risk score. (e) Concordance indexes of both clinical features and risk score.

Figure 8

Construction and a nomogram model. (a) A nomogram forecasts the ability of 1-, 3-, and 5-year OS of PC patients. (b) Calibration plot of the nomogram model.

Figure 9

Assessment of the immunotherapy reaction based on the m5C-LS in TCGA entire set. (a) Landscape of immune status between two groups. (b) GO enrichment analysis. (c) Difference of TIDE between two groups. (d, e) Waterfall plot of mutation values in the (d) high-m5C-LS group and (e) low-m5C-LS group. (f) Comparison of TMB between two groups. (g) K-M analysis based on the TMB. (h) K-M analysis combining TMB and the risk signature. ^∗p < 0.05, ^∗∗p < 0.01, and ^∗∗∗p < 0.001.