Linked SNPs and Monomorphic Sites in Observed SFS from RADseq Data

[mike...@gmail.com]

Hi All,

Does anyone have any advice/suggestions on how to deal with linked SNPs from short-read sequence data such as RADseq (i.e. multiple SNPs within a single locus) when assembling the observed SFS? If it's best to try to remove any linkage by, for example, retaining only the first SNP at each locus, then should the number of monomorphic sites be scaled down in any way - maybe by the proportion of SNPs that were retained/unlinked?

[Laurent Excoffier]

Hi,

if you remove linked SNPs there is no way to scale down the number of monomorphic sites and you cannot use the mutation rate to get absolutetime or pop size estimates.

For parameter inference with in a model, linked SNPs are no problem. Linked SNPs would only be a problem for model testing, as the likelihood woudl be a composite likelihood

best

laurent