A Qualitative Transcriptional Signature for Predicting Recurrence Risk for High-Grade Serous Ovarian Cancer Patients Treated With Platinum-Taxane Adjuvant Chemotherapy

Abstract

Resistance to platinum and taxane adjuvant chemotherapy (ACT) is the main cause of the recurrence and poor prognosis of high-grade serous ovarian cancer (HGS-OvCa) patients receiving platinum-taxane ACT after surgery. However, currently reported quantitative transcriptional signatures, which are commonly based on risk scores summarized from gene expression, are unsuitable for clinical application because of their high sensitivity to experimental batch effects and quality uncertainties of clinical samples. Using 226 samples of HGS-OvCa patients receiving platinum-taxane ACT in TCGA, we developed a qualitative transcriptional signature, consisting of four gene pairs whose within-samples relative expression orderings could robustly predict patient recurrence-free survival (RFS). In two independent test datasets, the predicted non-responders had significantly shorter RFS than the predicted responders (log-rank p < 0.05). In a test dataset containing data for patient pathological response state, the signature reclassified 12 out of 22 pathological complete response patients as non-responders and two out of 16 pathological non-complete response patients as responders. Notably, the 12 predicted non-responders in the pathological complete response group had significantly shorter RFS than the predicted responders (log-rank p = 0.0122). This qualitative transcriptional signature, which is insensitive to experimental batch effects and quality uncertainties of clinical samples, can individually identify HGS-OvCa patients who are more likely to benefit from platinum-taxane adjuvant chemotherapy.

A risk model of gene signatures for predicting platinum response and survival in ovarian cancer

“…Liu et al used the TCGA dataset to validate a seven genes-based model which can predict the survival of FIGO stage IIIc serous ovarian carcinoma (HG3cSOC) and served as a valuable marker for the response to platinum-based chemotherapy [33], however, they chose HG3cSOC to analysis and confined to gene expression analysis only; Zhao et al identified AGGF1 and MFAP4 as potential predictors of primary platinum-based chemoresistance [34], nonetheless, they just focused on the gene expression level and lack of experimental-level validation, such as immunohistochemistry; Dugo et al mainly focused on HGSOC patients who received complete cytoreduction (R0) and analyzed focal copy number alterations [35]; A qualitative transcriptional signature for predicting recurrence risk for high-grade serous ovarian cancer patients was constructed by Liu et al [36]; Salinas et al found 19 SNPs were associated with chemoresponse [13]. Although these studies have similarities to ours, we conducted a comprehensive analysis from diversified levels to figure out the most robust markers indicating platinum treatment response and prognosis.…”

Identification of prognostic genes after platinum-based chemotherapy in ovarian cancer

“…After optimization of the prediction model, a 34-gene model improved performance, and the AUCs could approach 0.80 [22]. And Liu and his partners developed a qualitative transcriptional signature based on 226 samples of highgrade serous OC patients receiving platinum-taxane adjuvant chemotherapy in the TCGA dataset, to predict patient recurrence-free survival [23]. Similarly, Lu and his members found a 10 prognostic gene panel to predict recurrence-free survival and overall survival in OC patients by the machine-learning method [24].…”

Predict ovarian cancer by pairing serum miRNAs: Construct of single sample classifiers

“…Studies have demonstrated this biological coordination phenomenon, whereby REOs of genes are broadly stable in normal samples and altered when a disease such as cancer occurs ( 10 , 13 ). More importantly, REOs have the unique advantage of being insensitive to batch effects, data normalization methods, partial RNA degradation, RNA amplification bias, and the proportion of different cancer epithelial cells ( 14 , 15 ). REO-based SSCs can therefore be used as diagnostic biomarkers for cancer and are particularly suitable for individual clinical diagnosis.…”