SEQanswers (
-   Sample Prep / Library Generation (
-   -   Brainstorming "ddRAD-BiS" (

lukeSeq 10-19-2015 07:40 PM

Brainstorming "ddRAD-BiS"
Hi there,

I'm a newbie with sequencing, and I'm currently thinking about new ideas for a grant application. I work on non-model organisms with smallish genomes, and I'm interested in the causes and consequences of DNA methylation.

Existing work on DNA methylation in my field has used whole genome bisulphite sequencing, which is expensive to do at high coverage, so past work often has zero biological replication. Alternatively, people have used poorly repeatable and low resolution methods, like methylation-sensitive AFLP or densitometry to get at the general 'vibe' of methylation in a species.

I had an idea involving ddRAD that I think might be great, though I may be missing something crucial (apologies for any serious newbie mistakes). I'd be keen to hear your thoughts on the idea, and suggestions regarding improvements and specifics. Here's the rough protocol:

1. Double digest the genomic DNA from one individual.
2. Add methylated adapters then do size selection.
3. React the sample with BiS - the unmethylated parts of the genomic DNA in each RAD marker will be converted, but the methylated adapter parts will not. Alternatively, use adapters where it doesn't matter if the Cs become Ts in some of the adapter sequences (e.g. it will still stick to the flowcell).
4. Sequence with e.g. HiSeq. Aim for higher coverage than usual in ddRAD work since we are looking for methylation, and so we need consistent high coverage across individuals to make good inferences.
5a. Idea 1: Sequence some of the DNA from a handful of individuals without BiS conversion. Use this as a reference to identify methylation (i.e. places where there is a C in the reference and a T in the sample).
5b. Idea 2: Sequence every sample twice: half the library is BiS-treated and half is non-BiS. Then for each sample, use the non-BiS sample as a reference. Less room for error than idea 1?

The outcome is hopefully:
1. Identification of methylated loci in randomly distributed bits of the genome for lots of individuals for minimal cost
2. Identification of methylation marks that vary between individuals
3. Simultaneous gathering of SNP and methylation data on the same set of individuals
4. Room for fun analyses: build a phylogenetic tree based on methylome similarity (compare populations/species/ecotypes/treatment groups); ask whether genetically variable regions are more diverse in methylation status, and whether they have more methylation; get a very precise estimate of the proportion of methylated loci in the genome in lots of species (likely more accurate than existing methods for doing this, e.g. HPLC, densitometry)

Thanks for reading!


nucacidhunter 10-19-2015 09:43 PM

This idea is very similar to representation reduced bisulphite sequencing (RRBS) and there are papers describing single enzyme and double enzyme digestion versions with or without size-selection.

lukeSeq 10-19-2015 10:27 PM

It sure is. Here's a Wikipedia on RRBS for interested readers

I guess you mean this paper - the that's the only paper I could find using double digest and BiS.

Still keen for thoughts on the specifics of the design, but it does like the idea is workable! Lots of things one can do with joint SNP and methylation data for a bunch of alleles/individuals.

kerplunk412 11-02-2015 01:51 PM

The general idea sounds good to me but I do have some questions/suggestions:

How small is "smallish" for the genome? I guess you have probably already done the math but it is surprising how few reads are actually needed for reasonable coverage of a small genome.

Do you expect methylation to mainly be at CpGs? I believe the idea of RRBS is that digestion by MspI enriches CpGs. Any reason to deviate from this approach for your experiments?

If the genome you are working with hasn't already been sequenced it may be good to sequence one or a few individuals first so you can create an in-silico map of your digestion, which may help you choose enzymes that will best achieve the desired coverage.

You can buy a kit to create your libraries. The company I work for, Bioo Scientific, sells one, and I believe Illumina and one or two other companies have kits as well.

Edit: I realize my last statement could have been clearer. I meant that you can buy a kit specifically for bisulfite-seq library prep. Feel free to PM me for more information.

lukeSeq 11-07-2015 06:03 PM

Hi Kerplunk,

Thanks for that reply, much appreciated! The genomes I'd be working on are around 300MB. I guess I could afford to sequence quite a few whole genomes with respectable coverage, but I'm interested in getting the costs down low enough to do masses of biological replication (it's amazing how many studies have 1 or a couple of replicates - surely this opens the door to artefacts that happen to look like biology, e.g. incomplete BiS conversion in sample A but not sample B, not to mention the fact that A and B are differentially methylated purely by chance?).

I do think methylation will be mostly in CpGs (like CCGG for MspI), so you're right that it'd be good to use a cutter that enriched for these regions, just as proposed in RRBS. Thus I was thinking of cutting with MspI and some other generic, non-methylation sensitive cutter, possibly after doing some in-silico trials in the genomes of close relatives.

p.s. for interested readers, have a look at this new methods paper which is very similar to what I had in mind, yet slightly distinct. Doesn't use BiS treatment, just cuts with a methyl-sensitive enzyme, and looks for RAD tags that drop out in some individuals

All times are GMT -8. The time now is 05:54 PM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2020, vBulletin Solutions, Inc.