What is the differentce between the folded and unfolded SFS input obo input file?

[gaoji...@gmail.com]

Dear all,

I have a question about the input obs.file for the fastsimcoal simulation.
What is the difference when you using the folded and unfolded SFS in fastsimcoal?
When I prepare the SFS input file in Arlequnin, I have vcf file bue the ancestral state is not knowen, I only got the MAF.obs output. But from the Arlequin manual said the both derived and minor allele SFS are produced in all cases, even if the ancestral state is not known. So I don't know the problem is?
I am appreciate if anyone could answer my question, I am stuck here.

Thank you very much!

[Tony Kess]

Hi,

The unfolded SFS refers to a SFS when the ancestral allele is known, and all alternative loci are assumed to be mutated since divergence from ancestral state - in Arlequin this allele is coded as "0". My understanding is that if that "0" allele is missing, the derived SFS won't be calculated. If the "0" allelic code is included but is not based on the true ancestral allele, the derived SFS WILL be calculated, but it will be wrong if you arbitrarily picked an allele as "0" instead of ensuring that allele is the ancestral version. I believe that's the scenario of derived SFS calculation the Arlequin manual is referring to, and cautioning against.  

Also, if you don't know the ancestral state, I think that you DON"T WANT the derived SFS because it will not be correct. I think it is better in your case to calculate the folded minor allele SFS instead, as Arlequin is currently doing. 

Hopefully this is correct, based on my understanding from using FSC for a month or so.

Tony 

[Laurent Excoffier]

Yes Tony is right.

If the ancestral allele is not known, the DAF SFS (unfolded spectrum) is irrelevant and all estimations should be done on the folded spectrum

best

laurent