*4.1. Transcriptome Data Assembly*

Log2 transformed RNA sequencing data generated by IlluminaHiSeq (Illumina, San Diego, CA, USA) and publicly available by the TCGA Research Network were downloaded via the UCSC Xena browser (http://xena.ucsc.edu, PCa *n* = 497, plus normal adjacent kidney tissue (NAT) *n* = 52; Table S1) [4,5].

Microarray data (Affymetrix Human Genome U95C Array; Affymetrix, Santa Clara, CA, USA) from the first prostate cancer progression cohort for DONSON, KI67, and AR were downloaded via Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/, GSE6919) [15]. The expression profiles of 25 androgen-deprivation resistant metastatic samples derived from four patients were obtained from different metastatic sites and were thereby used as individual samples (pPCa *n* = 66, Met(CRPC) *n* = 25). Normalized log2 mRNA (DONSON, Ki67, AR) expression data and the clinical features of the second investigated progression cohort were obtained from http://cbio.mskcc.org/cancergenomics/prostate/, which included primary PCa and metastatic samples (GSE21032, pPCa *n* = 131, Met *n* = 19) [14].
