Re: MethPipe DMR + HMR

[Andrew D. Smith]

Hi Dorina.

I'm forwarding to the group to see if someone can answer this.

Cheers,

Andrew

On Apr 7, 2014, at 2:23 AM, Lenz, Dorina <dl...@izw-berlin.de> wrote:

Dear Andrew,

I am using the MethPipe software and have a few questions regarding the computation of the HMRs which are then be used to calculate the DMRs.

According to the MethPipe software manual, it is recommended to perform the hypo-methylation calling if dealing with mammals. So, I have performed the normal hmr computation.

What exactly is the difference in computation if performing the hyper-methylation calling for plants?
Is the algorithm comparable?
If we have two groups, one control and one treatment.
If the control group has a hypo-methylated region, can we say then something about the hyper-methylation? E.g. will the treatment group have a hyper-methylated region for these positions?


Best Regards,
Dorina Lenz

Bioinformatician
Evolutionary Genetics

Leibniz Institute for Zoo and Wildlife Research (IZW)
in the Forschungsverbund Berlin e.V.
Alfred-Kowalke-Straße 17
10315 Berlin
Germany

Tel.:  +49 (0) 30 5168340
Fax.:  +49 (0) 30 5126104

www.izw-berlin.de

[Ben Decato]

Hi Dorina,

MethPipe provides two ways to identify hypermethylation in plants.  The first way is to invert the methylation levels and use the regular two-state HMM (the 'hmr' program).  If you are calling hyperMRs in this manner, it is safe to call the regions in your mammalian treatment set not identified as HMRs as "HyperMRs".  Just keep in mind that the majority of the genome is highly methylated in mammals, and therefore it makes more sense (and is more interesting) to describe the loss of hypomethylation in the treatment set than it does to describe hypermethylation at the site.

If you use the hmr-plant method to identify hyperMRs then the two are not directly comparable, since hmr-plant implements a 3-state HMM tailored to features specific to plant methylation.

Let me know if you have any other questions!

Ben

--
You received this message because you are subscribed to the Google Groups "MethPipe and MethBase Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to methpipe+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
-- 
Benjamin Decato
PhD Student
Computational Biology Section,
Department of Biological Sciences,
University of Southern California
1050 Childs Way, RRI 408A
Los Angeles, CA 90089-2910

[Andrew D. Smith]

I think the missing point here is that

(1) plants have default low methylation through the entire genome, punctuated with short regions of high methylation

(2) mammals have default high methylation everywhere

If you try and "identify" a region of low methylation in plants, you will basically find the whole genome, which isn't useful.

Similarly, if you try to identify a region of hypermethylation in mammals, you will find the whole genome. Also not useful.

Probably you want DMRs if you are looking to compare treatment vs. control, for example.