BiasCorrector: Fast and accurate correction of all types of experimental biases in quantitative DNA methylation data derived by different technologies.

Quantification of DNA methylation in neoplastic cells is crucial both from mechanistic and diagnostic perspectives. However, such measurements are prone to different experimental biases. Polymerase chain reaction (PCR) bias results in an unequal recovery of methylated and unmethylated alleles at the sample preparation step. Post-PCR biases get introduced additionally by the readout processes. Correcting the biases is more practicable than optimising experimental conditions, as demonstrated previously. However, utilisation of our earlier developed algorithm strongly necessitates automation. Here, we present two R packages: rBiasCorrection, the core algorithms to correct biases; and BiasCorrector, its web-based graphical user interface frontend. The software detects and analyses experimental biases in calibration DNA samples at a single base resolution by using cubic polynomial and hyperbolic regression. The correction coefficients from the best regression type are employed to compensate for the bias. Three common technologies-bisulphite pyrosequencing, next-generation sequencing and oligonucleotide microarrays-were used to comprehensively test BiasCorrector. We demonstrate the accuracy of BiasCorrector's performance and reveal technology-specific PCR- and post-PCR biases. BiasCorrector effectively eliminates biases regardless of their nature, locus, the number of interrogated methylation sites and the detection method, thus representing a user-friendly tool for producing accurate epigenetic results.

GHSR DNA hypermethylation is a common epigenetic alteration of high diagnostic value in a broad spectrum of cancers.

Identification of a single molecular trait that is determinant of common malignancies may serve as a powerful diagnostic supplement to cancer type-specific markers. Here, we report a DNA methylation mark that is characteristic of seven studied malignancies, namely cancers of lung, breast, prostate, pancreas, colorectum, glioblastoma and B cell chronic lymphocytic leukaemia (CLL) (n = 137). This mark was defined by substantial hypermethylation at the promoter and first exon of growth hormone secretagouge receptor (GHSR) through bisulfite pyrosequencing. The degree of aberrant methylation was capable of accurate discrimination between cancer and control samples. The highest sensitivity and specificity of cancer detection was achieved for cancers of pancreas, lung, breast and CLL yielding the area under the curve (AUC) values of 1.0000, 0.9952, 0.9800 and 0.9400, respectively. Narrowing to a single CpG site within the gene's promoter or four consecutive CpG units of the highest methylation levels within the first exon improved the detection power. GHSR hypermethylation was detected already at the early stage tumors. The accurate performance of this marker was further replicated in an independent set of pancreatic cancer and control samples (n = 78). These findings support the candidature of GHSR methylation as a highly accurate pan-cancer marker.