FST Structure

[SADS]

Hi,

I am using structure to analyse SNPs from whole genome sequence data.

I have run structure with a burnin period of 10,000, and MCMC 10000
using the admixture and linkage model.
I ran 10 times for each K (1 to 3) and I compared the ln Pr(K|X) as
described in Pritchard et al (2000). I found the value K = 3 as I was
imagining.

The triangle plot showed me 3 different groups, each one in a opposite
corner. But I found FST = 0.16, meaning small difference between the 3
populations. I was hoping a higher value of FST.

I would be very grateful of any suggestions on how to analyse this
problem.

Thanks
Ricardo

[Marco Pessoa]

Hi Ricardo,

I have found overall Fst values of 0.17 in some of my analyses, which some references I checked regarded as a moderate differentiation between groups, so I don't think your value would mean a small difference.

I would advise checking the values between groups, and not only the overall Fst value. That is, Fst values between groups 1 and 2, 2 and 3, and 1 and 3. 

If you still feel your obtained values are unexpected, run your group differentiation analysis on a different software and see if Fst values are still the same between inferred groups.

In my case, working with SSR markers, I did that using both Structure and Powermarker.

I have another suggestion: since you expected K = 3, I would run analyses for higher numbers of K (1 to 6, for example, so that you have a better view of how ln Pr (K|X) behaves).

Good luck,

Marco Pessoa-Filho

Researcher - Molecular Biology, Genetic Resources

Embrapa Cerrados

Brasilia - DF

Brazil

--
You received this message because you are subscribed to the Google Groups "structure-software" group.
To post to this group, send email to structure...@googlegroups.com.
To unsubscribe from this group, send email to structure-softw...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/structure-software?hl=en.

[SADS]

Hi Dr. Marco Filho,

I did as you commented about the number of k (1 to 3 and after 1 to 5). I found the best value of ln in k=3.
I'm going to check PAIRWISE.FST and test the PowerMarker software.

Thanks for your comments.

Cheers,

Ricardo