Research Paper Volume 15, Issue 19 pp 10322—10346

Figure 6. Feature gene selection. (A, B) RandomForest error rate versus the number of classification trees. (C) Biomarker signature gene expression validation by support vector machine recursive feature elimination (SVM–RFE) algorithm selection. (D) The intersection genes of SVM-RFE and RandomForest were screened by Venn diagram. (E) The genes included in our model were intersected with key characteristic genes to obtain PTGIS. (F) The ROC curve of PTGIS predicted the incidence of UCEC in TCGA database. (G) The ROC curve of PTGIS predicted the incidence of EC in GSE17025. (H) Box plots showed the expression of PTGIS in normal and UCEC tissues from TCGA. (I) The transcription levels of PTGIS in UCEC compared with the paired normal endometrial tissue was showed based on TCGA datasets. (J) Box plots showed the expression of PTGIS in normal and EC tissues from GSE17025. (K) Expression of PTGIS in pan-cancer.