Protein-coding genes, long non-coding RNAs combined with microRNAs as a novel clinical multi-dimension transcriptome signature to predict prognosis in ovarian cancer

Abstract

Ovarian cancer is prevalent in women which is usually diagnosed at an advanced stage with a high mortality rate. The aim of this study is to investigate protein-coding gene, long non-coding RNA, and microRNA associated with the prognosis of patients with ovarian serous carcinoma by mining data from TCGA (The Cancer Genome Atlas) public database. The clinical data of ovarian serous carcinoma patients was downloaded from TCGA database in September, 2016. The mean age and survival time of 407 patients with ovarian serous carcinoma were 59.71 ± 11.54 years and 32.98 ± 26.66 months. Cox’s proportional hazards regression analysis was conducted to analyze genes that were significantly associated with the survival of ovarian serous carcinoma patients in the training group. Using the random survival forest algorithm, Kaplan–Meier and ROC analysis, we kept prognostic genes to construct the multi-dimensional transcriptome signature with max area under ROC curve (AUC) (0.69 in the training group and 0.62 in the test group). The selected signature composed by VAT1L, CALR, LINC01456, RP11-484L8.1, MIR196A1 and MIR148A, separated the training group patients into high-risk or low-risk subgroup with significantly different survival time (median survival: 35.3 months vs. 64.9 months, P < 0.001). The signature was validated in the test group showing similar prognostic values (median survival: 41.6 months in high-risk vs. 57.4 months in low-risk group, P=0.018). Chi-square test and multivariable Cox regression analysis showed that the signature was an independent prognostic factor for patients with ovarian serous carcinoma. Finally, we validated the expression of the genes experimentally.