OxyBS: estimation of 5-methylcytosine and 5-hydroxymethylcytosine from tandem-treated oxidative bisulfite and bisulfite DNA

Abstract

Summary: The use of sodium bisulfite (BS) treatment followed by hybridization to an Illumina Infinium BeadChip (HumanMethylation450 and MethylationEPIC) is a common method for interrogating 5-methylcytosine (5mC) at single nucleotide resolution. However, standard treatment of DNA with BS does not allow disambiguation of 5mC from an additional cytosine modification, 5-hydroxymethylcytosine (5hmC). Recently, it has been demonstrated that paired BS and oxidative bisulfite (oxBS) treatment on the same sample followed by hybridization to an Infinium microarray permits the differentiation of 5hmC from 5mC. Nevertheless, estimation of 5hmC and 5mC from tandem-treated arrays has been shown to produce irregular estimates of cytosine modifications. Results: We present a novel method using maximum likelihood estimation to accurately estimate the parameters of unmethylated cytosine (5C), 5mC and 5hmC from Infinium microarray data given the signal intensities from the oxBS and BS replicates. Availability and Implementation: OxyBS is an R package available on CRAN.

5-Hydroxymethylcytosine localizes to enhancer elements and is associated with survival in glioblastoma patients

“…We adapted the BS–oxBS technology to DNA from 30 glioblastomas and applied a novel algorithm, OxyBS, to obtain genome-wide 5 hmC and 5 mC estimates21. All samples were primary tumours, fresh frozen and IDH1 and IDH2 wild type.…”

“…To effectively analyze and interpret this data, we develop and apply a novel algorithm, OxyBS21, and produce the first epigenome-wide characterization of 5 hmC in glioblastoma. We demonstrate that the glioblastoma genome is globally depleted of 5 hmC and its distribution is dependent on genomic context.…”

Genome-wide abundance of 5-hydroxymethylcytosine in breast tissue reveals unique function in dynamic gene regulation and carcinogenesis

“…Study subjects were aged between 13 and 81 years (median = 54.5 years), weighed between 41-204 kg (median = 68 kg) and had BMI values between 15-63 (median = 25). To obtain total and single-nucleotide resolution estimates for 5hmC and 5mC we applied the OxyBS algorithm 29 to data from paired BS and oxBS treated DNA measured with 450K arrays. Altered expression of enzymes that process cytosine modifications on DNA (such as TET1, TET2, DNMT3A and DNMT3B) could potentially explain the observed variation in total 5hmC across samples, therefore we tested for correlation between total 5hmC and expression of enzymes that process cytosine modification.…”

“…To perform estimation of 5mC and 5hmC proportions in each sample, we applied the recently developed OxyBS algorithm, which uses a maximum likelihood-based method and more appropriate statistical constraints than previous methods to accurately predict methylated and unmethylated proportions. Code used to perform the estimation is available in R-package 'OxyBS' 29 . To identify CpG sites with greatest potential to be functionally relevant, we selected the 3876 CpGs among the top 1% median 5hmC value across all samples.…”

Reconciling Base-Pair Resolution 5-methylcytosine and 5-hydroxymethylcytosine Data in Neuroepigenetics

“…Biologically, 5-hmC is produced through direct oxidation of 5-mC (Shen et al, 2014), meaning 5-hmC β values are directly dependent on 5-mC β values. In addition to their biological relationship, 5-mC and 5-hmC β values generated from BS/oxBS experiments are also methodologically related, since calculation of 5-hmC is dependent upon either subtraction or a maximum likelihood estimation step (Booth et al, 2013;Houseman et al, 2016;Xu et al, 2016). Unless one were to measure 5-mC and 5-hmC directly through an alternative combination of the presented techniques, this methodological interdependence is unavoidable.…”