Skip to main content
Fig. 2 | BMC Cancer

Fig. 2

From: Gene expression signatures of neuroendocrine prostate cancer and primary small cell prostatic carcinoma

Fig. 2

Correlation profiles relative to meta-12 adenocarcinoma and NEPC centroids across datasets. Nearest centroid classification of NEPC datasets demonstrated NEPC sensitivities and specificities of 91% and 100% on training samples, 60% and 98% in SU2C, 80% and 100% in WCMC, and 63% and 94% in JHU-FFPE. Centroid correlation profiles were also evaluated for prostatic adenocarcinoma datasets (TCGA, MSK, Mayo-FFPE) and various human tissue or cell line datasets including SCLC (GSE43346), CCLE (cBioPortal), Human Body Index (GSD7307), ENCODE (GSE19090), and NIH Roadmap (GSE18927). Correlations were generally weaker in FFPE datasets (JHU-FFPE, Mayo-FFPE) and in WCMC derived primarily from biopsies. Rare outlier adenocarcinomas were present across datasets, usually related to low ARS. SCLCs generally had the most similar centroid profile to NEPC followed by small cell gastric carcinoma and CNS-related samples. In JHU-FFPE, 5 SCPCs appeared to cluster with adenocarcinomas, demonstrated ARS scores similar to adenocarcinomas (Fig. 1, Additional file 3: Figure S3), and are discussed further in the JHU-FFPE results section

Back to article page