Pairwise Fst or 95% CI for non-smoothed data

[Alex Hart]

Hi,

I am preparing a paper for submission and all of my work/analysis has been based on a non-kernel smoothed ref_map pipeline. This includes a lot of programs which use the locus catalog as input (such as BayeScan) so I would prefer not to repeat the pipeline with smoothing if possible as all post-stacks work will need repeating too.

One of my co-authors has requested that I supply p-values or 95% CI values for the pairwise population Fst data presented. I cannot seem to find this information anywhere in the ref_map or populations output. I quickly repeated the pipeline with the smoothing options enabled but the Fst values were slightly different to the original - and I still couldn't find the information I was looking for!

My questions are:
- Does Stacks produce this data somewhere, and I'm just missing it completely?

- Is there a way that I can calculate either the p-val or the 95% CI easily? From what I gather it involves bootstrapping but I am not experienced in these techniques. I tried to use the R package hierfstat but it kept crashing.

- This is more of a subjective thing, but is it necessary to report the significance for pairwise Fst data if it is not the focus of the paper? We are mostly concerned with the loci highlighted by genome scan techniques. 

Please help, I am a very lost and confused PhD student!!

Thanks in advance,

Alex

[Catchen, Julian]

Hi Alex,

 

For the pairwise SNP-based Fst values, the populations program provides a p-value and CI via a Fisher’s Exact Test for each nucleotide position. These can be found in the populations.fst_X-Y.tsv files, where X and Y are the pair of populations being compared.

 

julian

[Alex Hart]

Hi Julian,

Thanks for the quick response. Is there no overall (i.e. not per locus) p-value or 95% CI for the pairwise Fst? I assume I cannot just average all the p-vals/CI results :^)

Thanks,

Alex