Objective
Ovarian cancer is a gynecological malignant tumor with the highest mortality, it preferentially metastasizes to omental tissue, leading to intestinal obstruction and death. scRNA-Seq is a powerful technique to analyze tumor heterogeneity. Analyzing omentum metastasis of ovarian cancer at the single cell level may be more conducive to analyzing and understanding omentum metastasis and prognosis of ovarian cancer at the cellular function and genetic levels.
Methods
The scRNA-seq data of omentum metastasis site were acquired from the GEO (Gene Expression Omnibus) database, the data were clustered by Seruat package and annotated by SingleR package. Cell differentiation trajectories were reconstructed through monocle package, and obtained the marker genes of every branch. The microarray data of ovarian cancer were extracted from GEO and were clustered into omentum metastasis associated subpopulations according to the marker genes of single cell differentiation trajectories. The tumor microenvironment and immune infiltration differences between subpopulations were analyzed by estimate package and CIBERSORT. After that, based on the survival information and the intersection of bulk RNA-seq data of TCGA (The Cancer Genome Atlas) and marker genes of GEO subpopulations, univariate Cox and Lasso regression were performed to further establish an omentum metastasis-associated genes (OMAGs) signature, the signature was then tested by GEO data. Finally, the clinicopathological characteristics of TCGA-OV and GEO were screened by univariate and multivariate Cox regression analysis, in order to draw the nomogram.
Results
The scRNA-seq data (GSE147082) were clustered into 18 cell clusters, and annotated into 14 cell types. Reconstruction of differentiation trajectories divided the cells into 5 branches, a total of 833 cell trajectories related characteristic genes were obtained. Two microarray datasets (GSE132342 and GSE135820) were integrated to gain the expression data of 510 genes in 7846 patients, which were subtyped into 3 subpopulations by the cell trajectories related characteristic genes. K-M survival analysis showed that the prognosis of cluster2 was the worst and cluste1 was the best, p < 0.001. The tumor microenvironment showed that the ESTIMATE Score, Stromal Score in cluster 2 is significantly higher than that in the other two clusters, while Immune Score, Tumor Purity were significantly lower, p < 0.001; The immune infiltration analysis showed differences in the fraction of 10 immune cells among the 3 subpopulations, p < 0.05. Immune checkpoint analysis showed that there were significant differences in the expression levels of 6 genes in the 3 subpopulations, which were significantly correlated with prognosis. Subsequently, the scRNA-seq characteristic genes of 5 branches, GEO characteristic genes of 3 subpopulations were intersected with the bulk RNA-seq of 379 patients in TCGA-OV, and 11 candidate genes of OMAGs (omentum metastasis-associated genes) were obtained by Lasso regression. The composition of OMAGs was further refined by Cox multivariate regression analysis, and established a 6-OMAGs signature (RiskScore = 6.669*VCAN + 0.173*FN1–0.128*IGKC + 0.258*RPL21 + 2.133*UCHL1 + 0.457*RARRES1). Taking TCGA data as training set and geo data as test set, ROC curves were drawn to verify its sensitivity. Ultimately, 5 clinicopathological characteristics (age, tumor status, initial treatment outcome, stage, grade) were screened by univariate and multivariate Cox regression analysis, and the nomogram was drawn.
Conclusions
Finally, we constructed a clinical prognostic model related to omentum metastasis of ovarian cancer, which was composed of 6-OMAGs and 5 clinicopathological features.