The Predictive Values Of KCNN4 And S100A14 For Recurrence In Optimally Debulked Patients With Serous Ovarian Cancer

Background: Approximately 50-75% of patients with SOC(serous ovarian cancer)experience recurrence within 18 months after first-line treatment.Current clinical indicators are inadequate for predicting the risk of recurrence.In this study,we strived to excavate genes associated with recurrence in patients with SOC under the premise of optimal debulking surgery,and evaluate their predictive ability as predictors of recurrence in optimally debulked SOC patients.Methods: The curated Ovarian Data package was searched to screen and download microarray or RNA sequencing datasets for SOC patients.Signal-to-noise ratio(SNR)was used to excavate the differentially expressed genes between recurrent and non-recurrent patients in each selected dataset.Top 2000 genes which had high SNR in each dataset were selected and 2 intersected genes finally obtained,which were KCNN4 and S100A14.Fisher's exact test and log rank test were applied to measure the correlations of KCNN4 and S100A14 m RNA expression status with recurrence rates of optimal debulked SOC patients in all datasets.Recurrence prediction models were constructed by 4 different machine learning algorithms,that is,linear kernel support vector machine,radial kernel support vector machine,artificial neural network and random forest.KCNN4 and S100A14 centered interaction network was built based on STRING database.The shortest path between KCNN4 and S100A14 was extracted.The most probable regulation directions in this path were speculated by bayesian network inference based on the hill-climbing algorithm.Enrichment analysis of all genes involved in KCNN4 and S100A14 centered interaction network was performed by hypergeometric tests on GO,KEGG and REACTOME.Immunohistochemistry was used to detect the protein expression of KCNN4 and S100A14 in an independent clinic cohort of 127 optimally debulked patients with SOC.Log-rank test,univariate and multivariate Cox regression were performed to verify the independent prediction powers of KCCN4 and S100A14 protein expression status for recurrence in optimally debulked SOC patients.Results: After rigorous screening,7 datasets with totally 838 samples from 5 different platforms were enrolled in our study,including microarray dataset for ovarian cancer in TCGA,RNA sequencing dataset for ovarian cancer in TCGA,GSE17260,GSE26193,GSE30161,GSE49997 and GSE9891.In general,in these 7 datasets,higher m RNA expression status of KCNN4 or S100A14 was significantly correlated with higher recurrence rates of optimal debulked SOC patients in all datasets.The m RNA expressions of KCNN4 and S100A14 were significantly positively correlated.Except for individual cases,there were no obviously significant relationships between various clinical factors and m RNA expression levels/statuses of KCNN4 and S100A14.KCNN4 and S100A14 m RNA expression values may be regulated by DNA copy number variation(KCNN4: p=1.918e-05)and DNA promoter methylation(KCNN4: p=0.0179;S100A14: p=2.787e-13).Recurrence prediction models built in the TCGA dataset based on KCNN4 and S100A14 by linear kernel support vector machine showed the best prediction performance in the other 6 datasets(AUC:0.5442-0.9524).The KCNN4 and S100A14 centered interaction network primarily involved in potassium ion transport.The shortest regulation path between KCNN4 and S100A14 was identified,called the KCNN4-UBA52-KLF4-S100A14 axis.Our cohort supported that KCCN4 and S100A14 protein expression statuses were independent predictors for recurrence in optimally debulked SOC patients.Conclusions: The high expression of KCNN4 or S100A14 was associated with a high incidence of recurrence in optimally debulked SOC patients on both the m RNA and protein levels,in both multiple public datasets and our cohort data.This discovery might facilitate individualized treatment of SOC.