Construction and validation of a prognostic risk model for bladder cancer based on single-cell RNA sequencing
10.3760/cma.j.cn115355-20230317-00140
- VernacularTitle:基于单细胞RNA测序的膀胱癌预后风险模型的构建与验证
- Author:
Ziyuan WANG
1
;
Chun LIU
Author Information
1. 山西医科大学第一临床医学院，太原 030001
- Keywords:
Urinary bladder neoplasms;
Prognosis;
Medical informatics;
Single cell analysis;
Biomarkers
- From:
Cancer Research and Clinic
2023;35(9):685-692
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct and validate a prognostic model for bladder cancer based on single-cell RNA sequencing (scRNA-seq) bioinformatics analysis of prognosis-related differential expression genes.Methods:The bladder cancer scRNA-seq datasets like GSE135337 and GSE129845 were downloaded from Gene Expression Omnibus (GEO) database, and the data were updated in 2022 and 2019; the expression profile and the survival data of 165 bladder cancer samples in the conventional transcriptome dataset GSE13507 (the data were updated in 2020) were downloaded. Expression profile data of 414 bladder cancer samples and 19 paracancerous samples and clinical information of 405 bladder cancer patients were downloaded from The Cancer Genome Atlas (TCGA) database. R 4.1.2 software was applied in the quality control and downscaling clustering of 10 bladder cancer single-cell samples selected from the GEO database and the cell annotation was made. The cellular communication of single cell data in the GEO database was analyzed by using CellChat. Univariate Cox proportion hazards model was used to analyze the differential expression genes related to prognosis of bladder cancer. The prognostic risk model was constructed by using LASSO-Cox regression analysis and the risk score was calculated. According to the median risk score, the bladder cancer patients in TCGA database were treated as the training set and all patients were divided into high‐risk group and low‐risk group. GSE13507 dataset in GEO database was used as the validation set, and the Kaplan-Meier method was used to compare the overall survival of the two groups in the TCGA training set and the GEO validation set; the time-dependent receiver operating characteristic (ROC) curves were used to evaluate the predictive efficacy of the prognostic risk model. R 4.1.2 software was used to construct the nomogram for predicting the 1-, 3- and 5-year overall survival rates of patients. Correlation analysis of risk score and clinical characteristics of bladder cancer patients in TCGA dataset was performed. Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis and gene set enrichment analysis (GSEA) were performed.Results:In GSE135337 and GSE129845 datasets, a total of 50 263 cells were obtained after the filtration of quality control, including 43 519 uroepithelial cells. More interaction between uroepithelial cells and fibroblast could be found in the microenvironment of bladder cancer. Uroepithelial cells sent signals mainly through the midkine signaling pathway. Finally, 9 prognosis-related differential expression genes (SPINK1, FN1, EFEMP1, ELN, PCOLCE2, TUBA1A, COL14A1, TCF4, and TM4SF1) were screened and the prognostic risk model was constructed. The risk score was calculated as -0.019×SPINK1+0.028×FN1+0.025×EFEMP1+0.023×ELN+0.098×PCOLCE2+0.004×TUBA1A+0.047×COL14A1+ 0.004×TCF4+0.096×TM4SF1. Based on the median risk score (1.350), the overall survival of the high-risk group (≥1.350) was worse than that of the low-risk group (<1.350) in the training set and the valiation set. ROC curve analysis showed that the area under the curve (AUC) of 1-, 3- and 5-year overall survival rates in the training set and the validation set were larger than 0.65. Based on the age, staging and prognostic model risk score, a nomogram was constructed to predict the 1-, 3- and 5-year overall survival rates of patients, and its calibration curve was close to the ideal curve. The risk scores were elevated in patients aged more than 60 years old, M 1 in M staging, N 1, N 2 and N 3 in N staging, and stage Ⅲ and Ⅳ in TNM staging, and the differences were statistically significant (all P < 0.05) . Enrichment analysis showed that several significantly-enriched genes were associated with functions and pathways such as humoral immune response, granulocyte chemotaxis, cytokine-cytokine receptor interactions, and B-cell-mediated immunity. Conclusions:The stable prognostic prediction model for bladder cancer constructedbased on scRNA-seq data can provide a reference for clinical assessment of patients' prognosis.
