meDIP-chip/meDIP-seq/MethylCap-seq using mouse DNA

[Jones90]

Hi all,

I am begining to analyse global methylation using my balb/c mice DNA.

I have been looking into some techniques and have came across; meDIP-CHIP, me meDIP-seq and MethylCap-seq.

Does anyone have any experiance regarding these techniques and mouse DNA.

Would anyone know which technique would be better for cost as I have a large number of samples to test.

Many thanks
B

[simonandrews]

It would be worth considering bisulphite based methods in addition to the enrichment methods you mentioned. We used to do a fair bit of MeDIP but have completely moved over to BS-Seq now (the systems we work in were particularly unsuited to MeDIP which helped spur this decision). Whilst whole genome bisulphite may be prohibitively expensive if you have a large number of samples RRBS or a pre-capture(the link is for human, but I think a mouse version is in the works) of interesting regions followed by bisulphite might be feasible.

The enrichment methods you mentioned work OK if you're looking for relatively small scale changes in an otherwise constant background - but their analysis and interpretation isn't simple and you can get misleading results. Bisulphite methods make the initial data generation more expensive, but the subsequent analysis and interpretation is much more straight-forward.

Another option you might want to look into if you want a cheap and large scale method would be the methylation array systems (as opposed to MeDIP-ChIP), which we've not used ourselves, but from which I've seen a few large datasets where these seemed to have worked well.

[dpryan]

Our experience seems to be similar to Simon's. We have one project using hybrid mice for which we have MeDIP-chip, MeDIP-seq, and RRBS data. I have to say that the RRBS is proving to be the most informative. As Simon mentioned, the MeDIP stuff becomes difficult to deal with unless you're looking for something in particular or are expecting large changes. Depending on the nature of your experiments and what you have access to (e.g., if, unlike us, you're near a core sequencing facility then you have more reasonable options), you might be able to do RRBS (or whatever you choose) on a small number of samples and use the others for validation. That sort of approach might be able to keep your costs under control while still keeping the resulting data quality high.

[ETHANol]

I'll be the third to basically say the same thing. In my experience MeDIP-seq data is really messy and can be a bit hard to interpret. Not impossible but the data analysis probably going to be complicated. RRBS-seq seems enrichment seems a little too random but it does work. Targeted enrichment, while more expensive, I think is the way to go at this point. While I only have first hand experience with MeDIP-seq but this is what I think is the best path.

If you still want to give MeDIP-seq a go, the protocol I've been using works really well and is here:

http://ethanomics.wordpress.com/medip-seq-protocol-with-truseq-adapters/

[fabified]

hello!
bisulphite sequencing is not a state of the art technique, since it also detects 5-hydroxymethyl-cytosine, which does not resemble the same function as normal methyl cytosine. I would suggest you to do MeDIP-seq.

[simonandrews]

hello!
bisulphite sequencing is not a state of the art technique, since it also detects 5-hydroxymethyl-cytosine, which does not resemble the same function as normal methyl cytosine. I would suggest you to do MeDIP-seq.

It's certainly true that BS-Seq does detect both mC and hmC. If you really want to see just mC then you could do oxBS-Seq, but the protocols for that are fairly new, and it's a lot harsher on your material in its current form. Under pretty much every condition we've seen hmC is always a tiny fraction of mC so a BS-Seq library is a very good approximation to the mC levels. You also don't see hmC at places where mC is not present.

MeDIP also has its limitations. In particular the signal you get from it can't be split into different C contexts, so you can't figure out the relative contribution of methylation in CpG CHH and CHG, and it's poor at detecting global changes in methylation.

There are pros and cons to each of the available techniques, but having had to deal with all of the different kinds of data mentioned here I'd certainly be happiest to see another BS-Seq dataset appear on my desk as these have proved to be the ones from which we've seen the clearest and more reliable signals.

[fabified]

It's certainly true that BS-Seq does detect both mC and hmC. If you really want to see just mC then you could do oxBS-Seq, but the protocols for that are fairly new, and it's a lot harsher on your material in its current form. Under pretty much every condition we've seen hmC is always a tiny fraction of mC so a BS-Seq library is a very good approximation to the mC levels. You also don't see hmC at places where mC is present.

MeDIP also has its limitations. In particular the signal you get from it can't be split into different C contexts, so you can't figure out the relative contribution of methylation in CpG CHH and CHG, and it's poor at detecting global changes in methylation.

There are pros and cons to each of the available techniques, but having had to deal with all of the different kinds of data mentioned here I'd certainly be happiest to see another BS-Seq dataset appear on my desk as these have proved to be the ones from which we've seen the clearest and more reliable signals.

It certainly depends on what you are working on. But I guess you are right, since I can not comment on the reproducibility of BS-seq compared to MeDIP-seq or hMeDip-seq.