Whole transcriptome analyses identify pairwise gene circuit motif in serous ovarian cancer

Abstract

Background: Ovarian cancer is the most lethal gynaecological malignancy, resulting in approximately 14,000 deaths annually in the United States. Transcriptome data are emerging as an effective tool possibly leading to clinical applications for various cancers.

Methods: We collected eight serous ovarian carcinomas (SOCs) and eight normal ovary samples, and generated a whole transcriptome profile of human ovarian cancer using microarrays, including mRNAs, lncRNAs, and circRNAs. We constructed a competing endogenous RNA (ceRNA) network involving these three types of RNAs and identified immune-related circRNAs from the network. Moreover, we proposed a gene-pair filtering method to identify significant expression reversals from integrated multi-cohorts, which mitigates the technical variation and improves the statistical power.

Results: Three pairs of mRNAs (BIRC5:PRKCQ, PTK2B:OGN, and S100A14:NR2F1) were identified as promising biomarkers and were fused as an indicator (SOC index) for diagnostic prediction. Validation in three independent cohorts demonstrated that the SOC index carries a very high predictive capacity (average ROC, 0.99; sensitivity, 0.98; specificity, 1.00). Additionally, the SOC index exhibited its prognostic potential to discriminate SOC patients between early and late stage disease.

Conclusions: Our findings elucidate the repertoire of RNA expressions in SOC and identified three gene pairs for the primary screening of SOC. Further biological experiments of the three gene pairs are warranted in order to investigate the underlying function mechanisms involved in ovarian cancer development.