J Cancer 2020; 11(13):3751-3761. doi:10.7150/jca.44034

Research Paper

Improved Prognostic Prediction of Glioblastoma using a PAS Detected from Single-cell RNA-seq

Hongwei Liu1,2*, Qi Yang1,2*, Yi Xiong1,2,3, Zujian Xiong1,2,3, Xuejun Li1,2✉

1. Department of Neurosurgery, Xiangya Hospital, Central South University, Changsha, Hunan, 410008, China
2. Hunan International Scientific and Technological Cooperation Base of Brain Tumor Research, Xiangya Hospital, Central South University, Changsha, Hunan, 410008, China
3. Xiangya Medical School, Central South University, Changsha, Hunan, 410008, China
*Authors contributed equally to this work.

This is an open access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/). See http://ivyspring.com/terms for full terms and conditions.
Citation:
Liu H, Yang Q, Xiong Y, Xiong Z, Li X. Improved Prognostic Prediction of Glioblastoma using a PAS Detected from Single-cell RNA-seq. J Cancer 2020; 11(13):3751-3761. doi:10.7150/jca.44034. Available from https://www.jcancer.org/v11p3751.htm

File import instruction

Abstract

Glioblastoma (GBM) is a common malignant brain tumor of the central nervous system with a poor prognosis. In order to identify the prognostic signatures of GBM, we screened differentially expressed genes (DEGs) that were based on a single-cell RNA sequencing (scRNA-seq) dataset. These genes characteristically represent the intra-tumor heterogenicity of glioblastoma. Moreover, we performed univariate analysis, log-rank test and multivariate Cox regression analyses to confirm a gene set that could be related to the overall survival (OS) among DEGs. Prognostic associated signatures (PAS) were utilized to construct a model for predicting OS in GBM patients. When considering either the training or the validation sets, time-dependent receiver operating characteristic (ROC) curves all indicated that our model displayed an excellent predictive ability. Additionally, we analyzed PAS at the single-cell level and found that the PAS score was associated with somatic mutations and clinical factors. Three factors, which included the PAS score, radiotherapy status, and age, were all used to establish a nomogram to predict the 6-month and 1-year survival probabilities. In conclusion, we constructed an optimal model that was derived from scRNA-seq to better predict the survival probability of GBM patients. These genes might also act as potential prognostic biomarkers and enable surgeons to develop individually therapeutic schedules and improve the prognosis of GBM patients.

Keywords: glioblastoma, single cell, differentially expressed genes, survival analysis, prognostic model