Can you help me interpret these results?

[Pietro Delfino]

Hi everyone,

Can you help me interpreting the attached results?
According to Evanno I see K=2.
But also K=10 looking at L(K) mean +- sd, and the small peak in K x Delta K graph.
Also, is all that standard deviation normal from K=11 on?

Burnin 1x10^6
mcmc 1.5x10^6
10 runs for each K (1 to 20)

Side note: when I run the same but using 25 runs for each K, the small peak goes to k=9.

Thank you

[Pietro Delfino]

PS: How do I attach a file?

http://taylor0.biology.ucla.edu/structureHarvester/completedJobs/timeless-lily-0ec7/summary.html

[Pietro Delfino]

Another thing that I did not understand; when choosing the best K with clumpak, the best K by Evanno is of course still K=2, but looking at the Probability By K graph, it says "Using median values of Ln(Pr Data) the k for which Pr(K=k) is highest: 19 ".

http://clumpak.tau.ac.il/results.html?jobId=1492006336

[Pietro Delfino]

Ok now I can upload the files, sorry for disturbing with repeated messages.

[zeamne T]

Hi Pietro,

Do K=2 or K=19 or K=20 make biological sense for your species and populations? What do the structure bar plots at each of these K look like?

Since you have 2 peaks in your delta K plot, maybe you have 2 main groups, and also substructure within each of those 2 groups. Perhaps you could try running each of the 2 groups (delimited by K=2) separately, to see if there is structure within those groups?

Mean LnPD for K11-18 do seem a little high...but I'm curious about the earlier K values, e.g. K=2 where there seems to be no variation at all. Check that they started with different seed values.

Cheers,

YC

--
You received this message because you are subscribed to the Google Groups "structure-software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to structure-software+unsub...@googlegroups.com.
To post to this group, send email to structure-software@googlegroups.com.
Visit this group at https://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/d/optout.

[Pietro Delfino]

Dear YC

Thanks for your input.

To be honest, I don't know what to expect from this analysis; since I am dealing with a germplasm collection, we are talking about 617 genotypes of the same species and subspecies, all the individuals are genetically very close to each other. K=2 does not make biological sense; I would suspect that in this case structure gives K=2 only because it cannot test K=1. My idea is that there is panmixis (suppported by PCA). That's why I am more interested in minor peaks.

I don't have bar plots with me at the moment but I can tell you that individuals are very admixed for K=10 and greater.

As you suggest to run structure on each one of the K=2 subgroups it's a good idea, this is something they've already done on the same species. I'll try that.

Coming back to minor peaks, I tried to run DAPC and best K is 11 (close to the minor peak at 10 from structure). Would it make sense to validate structure K=10 subdivision with the DAPC result?

I've also tried to run BAPS and best K here is 18. This is also close to K=19 from clumpak Probability By K graph. Would it make sense to compare these two?

In the end my idea is that concerning the uppermost level of structure, there is not subdivision, but from K=10 on, many subgroups can be inferred.

Thanks again for you input, it's good to have a second view from the outside.

Any other suggestion is appreciated.

Best

To post to this group, send email to structure...@googlegroups.com.