Novice help: ddRAD for population genetics

[TonyaLander]

Hi all,

We're working with the plant genus Diascia. There's no reference genome, and there isn't data on the overall genome size. We want to use ddRAD for population genetic analysis (~15 populations of 30 individuals each) but we're not sure how to choose the fragment size/enzyme pair. Not much seems to have been done for population genetics with ddRAD in plants and so we have had trouble finding much useful precedent in the literature (most pop gen has been with fish and most plant ddRAD focuses on association mapping or linkage - maybe we've missed a key article?). From what we have gathered, it seems that for the best coverage and reliability of base calling, we would want around the 100-300 bp range for our fragment size. Any suggestions or guidance would be very welcome! Thank you!

[nucacidhunter]

I would suggest size selection in 250-350 bp range because:

1- wider range most likely will result in many tags requiring more sequencing
2- PCR bias toward smaller fragments that will affect evenness of coverage
3- will reduce presence of possible primer/adapter dimmers
4- will give a chance to sequence longer or paired end without sequencing adapters .

Generally a 6 and 4 cutter combinations are optimum for most species. So you may have to do a pilot with few combinations on a sample from each population, sequence the libraries and determine tag numbers and polymorphism level and choose the one that gives you required number of loci for your population analysis.

[SNPsaurus]

Most plants are heavily methylated in their repetitive regions. Because of this, a methylation-sensitive enzyme like PstI will cut mostly in the non-methylated genic regions. That genic region is usually in the few 100 million basepair range, so that is the range of genome size you should think about, not the potential gigabases of methylated junk.

If you are going to do ddRAD in a group of species, then a primary concern should be reducing locus drop-out because of the high genetic diversity. See:

http://onlinelibrary.wiley.com/doi/10.1111/mec.12276/abstract

Particularly examine Table 1, which shows the effect of a frequent-cutter on data loss and resulting bias in the genotyping. SNPs (and there will be plenty between species) will disrupt one of the two cut sites, and also create new 4-cutter sites in the fragment. Either of those events will potentially drop a locus from the size range selected.

Longer fragments will be more easily disrupted by the creation of new sites. If you don't think that is really an issue, consider that AAAA has BAAA,ABAA,AABA,AAAB, or 12 different 4bp sites that could be converted to the cut site by an A mutation. Non-palindromic 4-cutters might cut every 100 bp, so there could be a dozen such "almost sites" in each 100 bp of the fragment.

A possible countering design would be two 6-cutters (PstI - EcoRI perhaps) with at least one being methylation-sensitive. You add to the cut site target size (12 bp instead of 10 bp) but remove the frequency of site creation. You can then have the longer size fragments that nucacidhunter likes and you would probably want at least that size range as you'll create fewer fragments with two 6-cutters.

[nucacidhunter]

Probably the easiest way for pilot project would be to try different enzyme combinations using ezRAD with Pippin size selection so you do not have to spend funds or time designing different adapters.

In my experience two 6 cutter combinations results in larger fragments generally over 500 bp. ddRAD may not be the best approach if you are analysing multiple species because overlapping RAD-tags will be low but it is OK for one species.

Edit: https://peerj.com/articles/203/

[TonyaLander]

Thank you!

Thank you both for the helpful suggestions!