Although prostate cancer (PCa) is the second leading cause of cancer death among men worldwide, not all men diagnosed with PCa will die from the disease. A critical challenge, therefore, is to distinguish indolent PCa from more advanced forms to guide appropriate treatment decisions. We used Enhanced Reduced Representation Bisulfite Sequencing, a genome-wide high-coverage single-base resolution DNA methylation method to profile seven localized PCa samples, seven matched benign prostate tissues, and six aggressive castration-resistant prostate cancer (CRPC) samples. We integrated these data with RNA-seq and whole-genome DNA-seq data to comprehensively characterize the PCa methylome, detect changes associated with disease progression, and identify novel candidate prognostic biomarkers. Our analyses revealed the correlation of cytosine guanine dinucleotide island (CGI)-specific hypermethylation with disease severity and association of certain breakpoints (deletion, tandem duplications, and interchromosomal translocations) with DNA methylation. Furthermore, integrative analysis of methylation and single-nucleotide polymorphisms (SNPs) uncovered widespread allele-specific methylation (ASM) for the first time in PCa. We found that most DNA methylation changes occurred in the context of ASM, suggesting that variations in tumor epigenetic landscape of individuals are partly mediated by genetic differences, which may affect PCa disease progression. We further selected a panel of 13 CGIs demonstrating increased DNA methylation with disease progression and validated this panel in an independent cohort of 20 benign prostate tissues, 16 PCa, and 8 aggressive CRPCs. These results warrant clinical evaluation in larger cohorts to help distinguish indolent PCa from advanced disease.