Cancer cells display altered methylation signatures distinguishing them from normal cells. Originating from all tissues and cells of the body cell-free DNA (cfDNA) including aberrantly methylated DNA reflect epigenetic aberrations occurs not only in tumor cells but also in tumor microenvironment. Actually, aberrantly methylated cfDNA has proved to be a promising biomarker for noninvasive detection of cancer with several clinically-certified tests (Epi-pro Colon®, Epi-pro Lung®, Cologuard®). However only one test (Epi-pro Colon®) uses blood plasma – the most convenient source of cfDNA. Input of tissue and age specific methylation along with unidentified reasons lead to the presence of molecules with every conceivable cytosine-methylation patterns which decrease probability of tumor DNA identification. Among numerous variants of cfDNA methylation patterns only few reflect cancer related changes. To identify those tumor-specific profiles single nucleotide resolution of methylated cytosine locations in the individual circulating DNA molecules is obviously required. We performed target bisulfite sequencing of potential prostate cancer (PC) cfDNA markers (GSTP1; RNF219) isolated from blood plasma of 18 healthy donors (HD), 17 benign hyperplasia (BPH) and 20 PC patients using MiSeq platform (Illumina). RNF219 gene was shown to have high diagnostic potential in our previous comparative study of cfDNA from HD, PC and BHP patients using HCGI12k microarrays (Cortese, et al., 2012). GSTP1 is a common pathological DNA methylation event in PC and is most widely studied and promising methylation marker in the cfDNA of PC patients (Wu, et al., 2011). Selected loci were amplified after bisulfite conversion (Zymo Research) of cfDNA with methyl-independent barcoded primers and sequenced with coverage ranging from 23509 to 143953. Identification of CpG methylation status in DNA fragments was performed with the BiQ Analyzer HT Software. All statistical analysis was performed using R Statistical Software (version 3.1.1) To reveal diagnostically significant differences, several approaches to data analysis were used. Conventional approach is a prediction of patient’s diagnosis based on differences in methylation level of CpG-sites. Another approach used in this study relies on discrimination of cancer-related correlation between methylation statuses of CpG-sites within individual molecules of cfDNA – Intramolecular Correlation of Methylation Statuses (ICoMS). Study population was randomly subsampled into training and test cohorts. The logit regression model based on methylation level of CpG-sites achieved an area under the ROC curve (AUC) exceeding 0.94 in both cohorts for GSTP1 gene and 0.81 – for RNF219 gene. A novel approach to identify diagnostic significance of cfDNA methylation was based on comparison of correlation matrices (pairwise phi coefficient between methylation statuses of CpG sites) for HD, BHP and PC groups. Binomial regression models with LASSO penalization were used to predict patient’s diagnosis. ROC curves for fitted models show AUC, specificity and sensitivity as 0.99, 100% and 92% for GSTP1 gene and 0.99, 100% and 90% for RNF219 (test cohort). The ICoMS approach estimates diagnostic value of target genes and reveals cancer-related cytosine methylation. That could potentially input in gene expression, identify the features of tumor physiology and on a par with conventional approaches provide helpful information for rational design of noninvasive, methylation-specific cfDNA-based diagnostics for PC. Moreover, the planned further evaluation of relationship between the revealed correlations associated with cancer and tumor biology could increase the basic knowledge of carcinogenesis.