Changepoint detection in base-resolution methylome data reveals a robust signature of methylated domain landscape

Abstract

BackgroundBase-resolution methylome data generated by whole-genome bisulfite sequencing (WGBS) is often used to segment the genome into domains with distinct methylation levels. However, most segmentation methods include many parameters to be carefully tuned and/or fail to exploit the unsurpassed resolution of the data. Furthermore, there is no simple method that displays the composition of the domains to grasp global trends in each methylome.ResultsWe propose to use changepoint detection for domain demarcation based on base-resolution methylome data. While the proposed method segments the methylome in a largely comparable manner to conventional approaches, it has only a single parameter to be tuned. Furthermore, it fully exploits the base-resolution of the data to enable simultaneous detection of methylation changes in even contrasting size ranges, such as focal hypermethylation and global hypomethylation in cancer methylomes. We also propose a simple plot termed methylated domain landscape (MDL) that globally displays the size, the methylation level and the number of the domains thus defined, thereby enabling one to intuitively grasp trends in each methylome. Since the pattern of MDL often reflects cell lineages and is largely unaffected by data size, it can serve as a novel signature of methylome.ConclusionsChangepoint detection in base-resolution methylome data followed by MDL plotting provides a novel method for methylome characterization and will facilitate global comparison among various WGBS data differing in size and even species origin.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-015-1809-5) contains supplementary material, which is available to authorized users.

Establishment of diagnostic criteria for upper urinary tract urothelial carcinoma based on genome-wide DNA methylation analysis

“…In order to establish reproducible diagnostic criteria, we aimed at marker CpG sites possibly showing stable DNA methylation abnormalities during carcinogenesis. In general, most 5-methylcytosine residues seem to be controlled in a coordinated manner with their neighbours, rather than being independent, resulting in a domain wherein, all CpG sites show largely similar methylation levels [ 31 , 32 ]. Such stably regulated domains often include CpG islands.…”

LTR retrotransposons transcribed in oocytes drive species-specific and heritable changes in DNA methylation

“…Biological replicates were combined to increase resolution and coverage. Methylated domain landscape plots were generated using changepoint detection analysis 41 . For all PBAT and WGBS datasets (except Cast/Ei GVO PBAT), we calculated average DNAme over a given set of genomic coordinates with >4 CpGs with >5× coverage.…”

Parental experience modifies the Mimulus methylome

“…We constructed a RIL-94 reference genome, re-annotated genes (Additional file 1 : Table S1) and transposable elements (Additional file 2 Table S2), mapped our WGBS libraries to this reference using bmap [ 51 ], and used methimpute to impute methylation across the genome more accurately [ 52 ] (Additional file 3 : Methods S1).…”

“…Each vector was entered into the R package “changepoint” (2.2.2) using the pruned exact linear time (“PELT”) [ 47 ] algorithm and a manual penalty of 1.4. This more stringent penalty decreased false positives compared to the penalty of 1 used in the initial methylome PELT paper [ 51 ]. The PELT changepoint detection algorithm was designed to identify changepoints (here, in methylation difference between the two treatment groups) in large datasets where the computational demands increase only linearly with the number of observations.…”

“…The PELT changepoint detection approach was previously successful in fragmenting an individual methylome into “methyl-regions” [ 51 ]. Here we extended this approach to identify DMRs between groups of individuals.…”