For the screening step, genome-wide DNA methylation profiling data of 9 paired PCa and NPT samples (GEO identifier GSE89243), reported in our previous study [10], has been reanalyzed to identify potential PCa biomarkers. Samples were processed according to the manufacturer’s protocol using two-color Human DNA Methylation 1 × 244 K Microarrays, which interrogate 27,627 known CpG islands (Agilent Technologies, Santa Clara, CA, USA). Saturated, non-uniform, and outlier probe signals were excluded from the analysis. The Cy3/Cy5 fluorescence ratios representing methylated/reference DNA were calculated and normalized to obtain log2 values for further analysis. Probe annotations were uploaded from the SureDesign platform (; accessed 1 March 2018). They were updated using the UCSC Genome Browser (; accessed 1 March 2018) according to the latest version of the human reference genome (GRCh38). Probes that were not detected in ≥30% of all samples were filtered out. After additional group comparison-specific filtering, only probes detected in 100% of samples in at least 1 of 2 groups were compared. Fold change (FC) values were estimated, and a paired (where applicable) or unpaired t-test was applied. Calculations were performed with GeneSpring GX v14.5 software (Agilent Technologies).

The gene set enrichment analysis (GSEA) was performed using the publicly accessible online GSEA tool and Molecular Signatures Database (MSigDB, v5.2;; accessed 1 March 2018) [13]. The collection of 50 hallmark gene sets, which conveys a specific biological state or process and displays the coherent expression, were utilized for GSEA [14]. False discovery rate (FDR) q-value with the cut-off <0.05 was used to correct for multiple testing.