Multiple peaks for Delta K and unusual run

[confuse]

Hi,

I ran my project twice the second time with reduced population size from 39 to 35, the first time I received single peak at K=2 which is what I was expecting as well but the second time I got 2 peaks at k=2 & k=7 which was strange. when I looked carefully I saw that the structure ran for 4 replications at k=7 instead of set parameter 3, it ran 2 times for k=6.
could this be the problem of unusual extra peak at k=7?

Do I need to run again or I canm consider K=2 as the best K?

I have attached the delta K graphs from both the runs also a screen shot of structure run.

Please guide.

Thank you,

Massub

[Pablo G Goikoetxea]

Hi Massub

I do not think the double peak is caused by having 2 runs at K=6 and 4 runs at K=7. However, how did you manage to get the standard deviation of lnP(D) at K=6?

On the other hand, I have seen this type of double peaked graphs in my own data. My hunch is that they are caused by nested structures. However, it would be necessary to know more data (type of markers, differentiation among populations, lengths of the MCMC chains, Structure model, etc...) before giving a more concrete answer

Best wishes

Pablo

[Massub Tehseen]

Hi Pablo,

Thanks for your response, The markers are DArT, there are 35 genotypes and 986 markers, 100k MCMC chains and 100k burnin, admixture model, population is bread wheat.

A bit unrelated to the topic but I want to ask one more thing please If I may, I have obtained an association mapping using marker, trait and structure raw data and I could generate PCA by doing numerical imputation of genotypic data.

Now I cannot perform LD analysis as a result cannot obtain LD graphs.

I did the analyses on R-GAPIT package as well, and same I could not generate LD maps.

I am very new to this topic so please bear with me If my questions seem naive.

 

Thank you for your time and cooperation.

--
You received this message because you are subscribed to a topic in the Google Groups "structure-software" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/structure-software/_Tu5UAHZpXk/unsubscribe.
To unsubscribe from this group and all its topics, 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.

[Pablo G Goikoetxea]

Hi Massub

I think you should increase your runs (at least to be sure you do not have convergence problems). I would try 500K burning and 500K final runs.

If I recall correctly, DArT are dominant markers. Therefore, even if you can do a lot of analysis, please be aware many of them will be assuming Hardy-Weinberg equilibrium for all your markers. You should judge whether this might be a credible assumption or not.

On the other hand, in order to obtain LD maps you need to know the positions of your markers in the genetic linkage map (i.e., you need to genotype your markers in two parents and their progenies). I assume you do not know it, otherwise you would be using any of the Linkage models from Structure. Therefore, you can not do LD maps. However; you can still calculate LD among all the markers you wish (... and try to see if some markers show very high LD). There are many softwares that can do that (for example, Genalex, or Genepop, Fstat, etc)

Pablo

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

[Massub Tehseen]

Thank you so much for your suggestions. I'll try them and see what happens.

Thank you again.

Massub.

To unsubscribe from this group and all its topics, send an email to structure-software+unsubscribe@googlegroups.com.

To post to this group, send email to structure...@googlegroups.com.
Visit this group at https://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "structure-software" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/structure-software/_Tu5UAHZpXk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to structure-software+unsub...@googlegroups.com.

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

[Massub Tehseen]

Hi Pablo,

Finally I was able to do LD via R-GAPIT package. I have got both marker LD & density graphs but the LD graph seems a bit odd to me. I wonder is it ok.

If you be kind to please take a look and interpret whats going on.

Thank you,

Massub.

To unsubscribe from this group and all its topics, send an email to structure-software+unsubscribe@googlegroups.com.

To post to this group, send email to structure...@googlegroups.com.
Visit this group at https://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "structure-software" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/structure-software/_Tu5UAHZpXk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to structure-software+unsub...@googlegroups.com.

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

[Pablo G Goikoetxea]

Dear Massub

As I told you, you need to know the genetic distance among the markers in order to make an LD plot. 

Pablo

To unsubscribe from this group and all its topics, send an email to structure-software+unsub...@googlegroups.com.

To post to this group, send email to structure...@googlegroups.com.
Visit this group at https://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "structure-software" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/structure-software/_Tu5UAHZpXk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to structure-software+unsub...@googlegroups.com.

[Massub Tehseen]

Got it. Thanks a lot again.

To unsubscribe from this group and all its topics, send an email to structure-software+unsubscribe@googlegroups.com.

To post to this group, send email to structure...@googlegroups.com.
Visit this group at https://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "structure-software" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/structure-software/_Tu5UAHZpXk/unsubscribe.

To unsubscribe from this group and all its topics, send an email to structure-software+unsubscribe@googlegroups.com.

To post to this group, send email to structure...@googlegroups.com.
Visit this group at https://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "structure-software" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/structure-software/_Tu5UAHZpXk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to structure-software+unsub...@googlegroups.com.

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

[Pablo G Goikoetxea]

Independently of your graph, which should show the markers ordered by their positions on the genetic map (one graph per chromosome or even better chromosome arm), your plot shows some very large LDs. This could indicate close linkage, or...
Hope this helps
Pablo