Optimizing a ddRADSeq study with a large genome

[maa1024]

I am designing a study to investigate the population genetic structure of an endangered species. We've decided to go the ddRADSeq route, and I have a few questions for anyone who has experience with this sequencing method and/or population genetics.

My species has a large genome (~2Gb), and because I'm assessing change across the entirety of North America over a span of 100 years, I also have a large sample size (n=480).

In order to maintain a higher level of coverage (25x) while also remaining within the current budget, we can only target 0.25% (5 million bases) of the genome.

I was hoping for some insight on this issue. Although low, is 0.25% (and 5 million bases) of the genome per individual enough to answer questions regarding population genetic structure with that many individuals at that level of coverage? Or would it be prudent to reduce the amount of coverage in order to include more of the genome?

Relevant points: 1. No reference genome for this species, but there is one for a closely related species. 2. Very little is known about this species and no population genetics study has been conducted to compare to.

[Gopo]

If I understand you correctly, your samples could be up to 100 years old? Depending on the source of DNA, the DNA maybe sheared and low molecular weight, making it less amenable to ddRAD. If this is the case, one can do an initial ddRAD with high quality DNA samples and develop probes/baits to capture the variable RAD loci. Such target enrichment/sequence capture of the "RADnome" would be amenable for degraded DNA samples. See https://www.ncbi.nlm.nih.gov/pubmed/27416967 for more details on the idea of capturing the RADnome.

[maa1024]

Thank you for the suggestion! The age of a few of my samples is a concern, but they do not make up the bulk of my sample size. This is certainly something I should look into either way.

Do you have any suggestions regarding the balance between sample size, coverage, and percent of genome?

[Gopo]

Unfortunately, I can't really give too much advice about optimizing. Others on the forum have more experience than I do. I've personally subsampled a dataset of 16 individuals, 4 per population, genotyped at 20,000 SNPs down to ~100 random SNPs and have gotten very similar pairwise FST, observed and expected heterozygosity, allelic richness, and STRUCTURE results between the 20,000 SNP and 100 SNP datasets. But results will definitely vary, especially for species of conservation concern.