interpreting bar plot

[CR]

Hi Vikram,

I am attaching two barplots generated at different K's I am a bit confused about what they convey.

The top picture where there are two colours K=2 - does that mean no detected structure at this level or should we say there are two sub populations?

Bottom picture where there are 6 colours K=6 now every lines is mix of clusters and what does that mean? 

Sorry about this very basic question.

Regards

CR

[Ramesh krishnan Ramasamy]

Dear Chitrarag,

   Selecting suitable assumed population size (K) is the key in STRUCTURE analysis. However, many people like to have range of K values and discuss the results. Here, you used K=2 then jumped to K=6, what happened to 3-5? Any specific reason to use only 2 & 6? At K=2, your populations is more of structures along with good number of admixtures. At K=6, many of your individuals are admixtures. Try to answer these using biological nature of the sample.

Someone can improve my answer!

Regards

Ramesh

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

--

Ramesh krishnan R
Senior Research Fellow
Molecular Biology Lab-1
Central Sericultural Research & Training Institute
Ministry of  Textiles, Govt. of India
Srirampura, Manadavadi road
Mysore - 570 008

Mobile: +91 8951696997, +91 7502533038
Email: rame...@gmail.com

[CR]

Dear Ramesh,

Thanks. The run was for a range of K's and I just wanted to learn to interpret these barplots correctly. Yes these are related inbreds I would expect to see them more like admixtures. But does K=2 show that although they are admixtures they can be divide into 2 sub-populations as in K=2? As I move from k2-K6 they look more mixed

Thanks for your kind reply

CR

[Ramesh krishnan Ramasamy]

Dear Chitra,

 Obviously, the proportion of admixtures increases by increase in K. At K=2, you are getting both structured and admixtures. Whether to stick to 2 or more to be decided based on LnPD, Delta K and most importantly biology of the sample. Whether K=2 or more giving any biologically meaning to your data? I would like to recommend you to observe the each individual individually and compare with its biology of the sample like origin, pedigree and lot more. You can also infer K from other analysis like UPGMA and PCoA.

Regards

Ramesh

[CR]

Thanks Ramesh

[Vikram Chhatre]

Chitra -

What was the result of model complexity choice script?

Can you show us the barplots for K=3 through K=5 also?

V

[CR]

Dear Vikram,

Sorry to say these runs were not mine and I don't know the details I just got this image and wanted to understand it

Thanks CR

On Friday, 14 March 2014 14:21:13 UTC+8, CR wrote:

[Vikram Chhatre]

I see.  Can you cite the source of this figure?

Based on this figure, I would say the optimal K is either 2 or 3.  If K=3, first cluster is admixed with some membership in cluster # 2.  The second and third cluster show a nice gradual increase in membership of individuals as you move from left to right and vice versa.  

If you can produce bar plots from your own data, we can provide specific help.

V

[CR]

Interesting Vikram. Unfortunately this is not from a public source.

Would you say at K=2 the lines in the midddle are mixes of cluster 1 and 2 and therefore even at K=2 no structure is detected? As we move up to K6 all lines seem to be a mix of the represented clusters.

So what I am getting at is if I see a plot where the lines are a mix of the cluster as in K=6 could one say there is no distinct structure observed?

Thank you very much for helping out. I am running my own analysis and will see how that goes and discuss the same with you.

Regards

CR

On Friday, 14 March 2014 14:21:13 UTC+8, CR wrote:

[Vikram Chhatre]

Population structure is clearly evident in that data set, at both K=2 and K=3.  In fact the individuals in the last 1/4rd portion of the plot represent a distinct cluster through K=4.  

V

[CR]

Thanks Vikram - I think what you are pointing to are lines that belong to just one cluster ends of the plots K2-K4/5

Thanks for the help.

Regards

CR

On Friday, 14 March 2014 14:21:13 UTC+8, CR wrote: