Construction of a Novel Risk Model Based on the Random Forest Algorithm to Distinguish Pancreatic Cancers with Different Prognoses and Immune Microenvironment Features

Yalan Lei, Rong Tang, Jin Xu, Bo Zhang, Jiang Liu, Liang Chen, Qingcai Meng, Jie Hua, Xianjun Yu, Wei Wang, Si ShiResearch 2021 · DOI

Abstract

Abstract Background: Immune-related long noncoding RNAs (irlncRNAs) have been demonstrated to be actively involved in the regulation of immune status. With this study, we aimed to establish a risk model of irlncRNAs and further investigate the roles of irlncRNAs in prognosis prediction and the immune landscape in pancreatic cancer. Methods: The transcriptome profiles and clinical information of pancreatic cancer patients were retrieved from The Cancer Genome Atlas (TCGA). Immune-related genes (irgenes) downloaded from ImmPort were used to screen the lncRNAs by correlation analysis (R>0.5, p<0.001). Random survival forest (RSF) and survival tree analysis showed that 9 irlncRNAs were highly correlated with overall survival (OS) according to the variable importance (VIMP) and the minimal depth. Then, Cox regression was used to establish a risk model with 3 irlncRNAs, which was evaluated by Kaplan-Meier analysis. In addition, we performed Cox regression analysis to establish the clinical prognostic model, which showed that the risk score was the only independent prognostic factor (p<0.001). A nomogram was drawn to visualize the clinical features. Wilcoxon signed-rank test and analysis of the correlations between the risk score and immune cells was applied to explore the potential irlncRNAs related to immune status. Results: A total of 176 samples were randomly divided into a training set (n=123) and a test set (n=53). A total of 1903 irlncRNAs were identified after Pearson correlation analysis between irgenes and lncRNAs. A total of 9 irlncRNAs were identified in the survival tree analysis, and 3 irlncRNAs (LINC00462, LINC01887, RP11-706C16.8) were used to establish a risk model. The areas under curve (AUC) of the receiver operating characteristic (ROC) for 36 months, 30 months, 24 months, 18 months, 12 months and 6 months were 0.778, 0.774, 0.751, 0.753, 0.780 and 0.756, respectively, with a concordance index (C-index) of 0.696. In the clinical prognostic model, the C-index increased from 0.599 to 0.682 after combining the risk scores. Furthermore, a high risk was associated with increased infiltration of CD4+ T cells, M0 macrophages and M1 macrophages. Conclusions: We established a novel three-irlncRNA risk model. The clinical evaluation showed its robustness in predicting the prognosis and immune landscape of pancreatic cancer.
BibTeX
@inproceedings{lei2021construction,
  author = {Yalan Lei and Rong Tang and Jin Xu and Bo Zhang and Jiang Liu and Liang Chen and Qingcai Meng and Jie Hua and Xianjun Yu and Wei Wang and Si Shi},
  title = {{Construction of a Novel Risk Model Based on the Random Forest Algorithm to Distinguish Pancreatic Cancers with Different Prognoses and Immune Microenvironment Features}},
  booktitle = {Research},
  year = {2021},
  doi = {10.21203/rs.3.rs-454599/v1},
}

Loading related content...

Figure