Method to detect DMR

[serenaliao]

Is there any popular method to detect DMR(differential methylated regions) using Bisulfite sequencing data? I searched but didn't find many method except for some really naive ones. It seems there are more methods for differentially expressed genes in RNAseq.

Is there anyone who is familiar with this area to give some directions? Thanks!

[dpryan]

There's not a lot out there for bisulfite sequencing at the moment. There's a pipeline available using Seqmonk and there's also bsmooth, though I've not used it yet. Unfortunately, there's a lag between a technique catching on and the appearance of a lot of packages to analyze the resulting data (just look at the number of MeDIP-chip and MeDIP-seq analysis packages that came out in the last 2 years).

[serenaliao]

Thanks dpryan. I will look into those. If you are interested we can discuss this further.

There's not a lot out there for bisulfite sequencing at the moment. There's a pipeline available using Seqmonk and there's also bsmooth, though I've not used it yet. Unfortunately, there's a lag between a technique catching on and the appearance of a lot of packages to analyze the resulting data (just look at the number of MeDIP-chip and MeDIP-seq analysis packages that came out in the last 2 years).

[PeteH]

I'm familiar with BSmooth/bsseq combo. BSmooth is the alignment and post-processing suite (using Bowtie2 as the backend) and bsseq is the bioconductor package for smoothing methylation estimates and testing for DMRs. Here's the paper, the website for BSmooth (seems to be down at the moment) and the page for bsseq on Bioconductor. bsseq uses a local likelihood approach to smooth the raw methylation estimates at each CpG by exploiting the fact that nearby CpGs are similarly-methylated. Unlike many of the naive Fisher-type tests of differential-methylation at individual CpGs which ignore biological variability, the bsseq procedure makes use of this important information. The testing procedure is a little ad-hoc but seems to work fairly well.

Other Bioconductor packages that may be of interest (I haven't tried many of these):

• methVisual - designed for the older-style bisulfite sequencing, but may be useful/adaptable to your problem.
• methyAnalysis - looks interesting, I haven't seen this until now. From my brief read of the documentation it was originally designed for microarray data, but the authors seem to suggest some of the methods will be appropriate for BS-seq and that more functionality for sequencing-based studied of methylation are in the works. Does DMR testing using a sliding window approach that exploits the dependence of methylation at nearby CpGs.
• methylSeekR - A hidden Markov model for identifying regulatory regions from whole-genome BS-seq. Aims to identify partially methylated domains and lowly-methylated and unmethylated regions. Used in this paper, with a methods paper under submission.

Incidentally, I'm working on this topic in my PhD. Unfortunately I don't have any software to contribute at this time but hope to soon
Pete

[serenaliao]

Thank you Pete. The information is really useful.

I'm familiar with BSmooth/bsseq combo. BSmooth is the alignment and post-processing suite (using Bowtie2 as the backend) and bsseq is the bioconductor package for smoothing methylation estimates and testing for DMRs. Here's the paper, the website for BSmooth (seems to be down at the moment) and the page for bsseq on Bioconductor. bsseq uses a local likelihood approach to smooth the raw methylation estimates at each CpG by exploiting the fact that nearby CpGs are similarly-methylated. Unlike many of the naive Fisher-type tests of differential-methylation at individual CpGs which ignore biological variability, the bsseq procedure makes use of this important information. The testing procedure is a little ad-hoc but seems to work fairly well.

Other Bioconductor packages that may be of interest (I haven't tried many of these):

• methVisual - designed for the older-style bisulfite sequencing, but may be useful/adaptable to your problem.
• methyAnalysis - looks interesting, I haven't seen this until now. From my brief read of the documentation it was originally designed for microarray data, but the authors seem to suggest some of the methods will be appropriate for BS-seq and that more functionality for sequencing-based studied of methylation are in the works. Does DMR testing using a sliding window approach that exploits the dependence of methylation at nearby CpGs.
• methylSeekR - A hidden Markov model for identifying regulatory regions from whole-genome BS-seq. Aims to identify partially methylated domains and lowly-methylated and unmethylated regions. Used in this paper, with a methods paper under submission.

Incidentally, I'm working on this topic in my PhD. Unfortunately I don't have any software to contribute at this time but hope to soon
Pete

[shadow19c]

Hello,
I am wroking in A.Thaliana genome, I'm using BS-seq to detect Difference of methylation between different mutants.
I'm using Bismark for methylation call.

I want to know if there is a program to analyse DMR of different context (CHH, CHG, CHH) using this kind of table
Chr3 101 + 0 0 CHH CCC
Chr3 102 + 0 0 CHH CCT
Chr3 103 + 0 0 CHH CTA
Chr3 108 + 0 0 CHH CCC
Chr3 109 + 0 0 CHH CCT

Thanks