Just a clarification. In your input file, do you start by naming the haplotypes and giving the relevant sequence for each haplotype, or do you repeat each haplotype when you describe the populations?

I include every individual with its respective haplotype separately. So I do not define each haplotype at the beginning.

A possible explanation might be that when Arlequin calculates Fst from sequences also takes in account the distances between haplotypes. Something like Rst or Nst, depends on the literature. When you calculate the distances based on frequencies only, you probably consider each haplotype a separate entity and ignore any information about the differences among haplotypes. Is that the case?

If this is the case, then in your data you either have a strong phylogeographic structure, or your haplotypes are separated by some form of indel, which the software rates with a high distance among haplotypes. Or you have both.

Dear Aristotelis,

I chose the 'use haplotype frequency only' option. Under this option only the frequency and no other sequence property should be used. If I use the distance based options I do get different results, thus I am confident that distances were not calculated. In fact, I re-calculated my FST values using simple artificial haplotypes which differ in only two base pairs each and created an input file similar to my original one. And I do get the same strange outcome (different FST values for sequence based vs. frequency based input file)! Thus the results seem to be independent from the sequence properties and not to be an artifact of genetic distances!

Dear Martin,

now I am really curious and I am going to check previous analyses of mine and come back to you shortly. Have you tried to define haplotypes in the beginning? It would be interesting to see if this would change something. Actually it shouldn't!

Dear Aristotelis,

I attached two smaple files to my above question: one using frequencies the other using haplotypes. Both are derived from the same dataset. If you run them in  Arlequin you will see the difference in the Fst values (make sure that the option 'use haploytpe frequencies only' is really invoked, sometimes Arlequin calculates frequencies instead).

Dear Andrey,

I have been using the mitochondriall encoded COI. So there should be only a single gene variant (=haplotype) per individual.

Dear Martin, I have sent you a message asking for the input files. Have you seen it?

Dear Martin,

Please have a look into this article if you haven't seen yet. It compares different fixation indexes and distance estimators by simulating data types that differ in their genetic composition. May be your answer is here.