Fig. 4

PC-CIN separates primary tumors from CRPC metastases. a Heat map displays up-regulation of leading edge genes for CIN70 in M1-oligo and M1-poly tumors that significantly contribute enrichment of CIN, including seven genes highlighted in red, referred to as PC-CIN. b PC-CIN genes generated from a support vector machine (SVM) with a Gaussian kernel were determined to perform similarly to CIN70 in separation of PNBX case by M-stage with high accuracy. The receiver operating characteristic (ROC) curves showed the performance of the PCA-SVM classifiers using the CIN70 and PC-CIN gene signatures. c PC-CIN scores are significantly higher in metastases compared to primary tumors in mCRPC datasets GSE21034 and GSE35988 p. d Multidimensional scaling analysis of TCGA samples by PC-CIN scores (PC-CIN-High in red, PC-CIN-Low in blue). e Distribution of PC-CIN scores among TCGA PC cases stratified by Gleason sum (one-way ANOVA p < 0.0001) or (f) biochemical recurrence (rank-sum p < 0.0001)