Abstract: Objective The aim of this study was to draw single-cell transcriptome profiles of high-grade serous ovarian cancer(HGSOC),borderline ovarian cancer(OC),and normal ovaries in order to identify biomarkers that can diagnose and predict the prognosis of OC. Methods The differentially expressed genes between HGSOC,borderline OC,and normal ovarian tissues were analyzed using single-cell data sequenced(SRA database:PRJNA756768).The cell subsets associated with tumor progression were screened by functional enrichment,cell communication between different subsets was analyzed by Cellchat,and cell differentiation trajectories were explored by pseudotime analysis to finally determine the subsets most relevant to tumor progression.Combined with OC transcriptome data of OC from the Cancer Genome Atlas(TCGA)with patient prognosis,biomarkers for diagnosing and predicting survival of OC patients were ultimately screened. Results After using t-distribution stochastic neighbor embedding(t-SNE)for dimensionality reduction,nine cell subpopulations were obtained:endothelial cells,myeloid cells,fibroblasts,T cells,stromal cells,B cells,and 3 epithelial cell subpopulations(C1,C4,and C7).Further analysis revealed that copy number variation(CNV)in the C4 group had the highest score in HGSOC,higher than those of borderline OC and normal ovaries,and was negatively correlated with prognosis.DMKN was a key marker gene in this group.Transcriptome analysis of OC in the TCGA database showed a close correlation between DMKN and poor prognosis(P=0.026),and the diagnostic efficacy of DMKN for OC was significant(AUC=0.906). Conclusion This study is based on single-cell sequencing data to screen for DMKN,which can effectively diagnose and predict the prognosis of OC.This study provides new ideas for the diagnosis and prognosis prediction of OC.

Key words: Single cell sequencing, Ovarian cancer, DMKN, Prognosis