How to find suitable K value for determining genetic structure, I am using microsatellite markers.
I have used k from 1 to 10
Burnin Period 50000
MCMC rep 100000
How tocalculate delta K value using STRUCTURE
Dear Saurabh,
if you have a biological assumption about the underlying structure it could be justified to use this information. If you want to infer the number of clusters you could proceed as described in the STRUCTURE manual in chapter 5:
http://burfordreiskind.com/wp-content/uploads/Structure_Manual_doc.pdf
The linked papers nicely describe how to proceed in learning about the "optimal" cluster number using structure:
Article Detecting the number of clusters of individuals using the so...
Article An overview of STRUCTURE: Application, parameter settings, a...
Finally, you can also use a clustering algorithm such as e.g. k-means clustering and use the obtained cluster number as your working hypothesis. You could try out the R package adegenet with its DAPC function. It should also work with microsatellite data and has an implemented function based on k-means clustering for subpopulation identification.
But keep in mind that often a best k value is not available.
Kind regards
Patrick
Thank You@ Patrick Thorwarth for your valuable information and suggestion.
Sir @ Patrick Thorwarth I am not able to get Delta K . Do we need to calculate manually the Delta K. If Yes then How to calculate Ln (K)-Ln(K-1) I am not able to get.
I have Ln(K) values
-9700.5 -8351.8 -7458 -6924.4 -6477.6 -6010.8 -5679.7 -5354.2 -5048.3 -4668
Hello,
I agree with Patrick Thorwarth that you should go through the user’s manual for STRUCTURE. Generally, when you begin your experiment you surely have some hypothesis, assumptions, which you want to verify based on molecular data. Of course they should be built based on your knowledge concerning your research material.
Finally, in response to your question regarding calculation of delta K, I recommend you to use the STRUCTURE HARVESTER to analyze the output data from STRUCTURE.
http://taylor0.biology.ucla.edu/structureHarvester/
The STRUCTURE HARVESTER is user friendly online tool which help you to interpret the data received from STRUCTURE.
Good luck
Piotr
Dear Saurabh,
I never used STRUCTURE HARVESTER but it seems to be a well know tool.
You can calculate delta K on your own, using e.g. excel.
I attached you a toy example in excel (without warranty that I implemented the formulas correctly). The Run1 data points are the one you provided, the Run2 and Run3 I made up. I think it is important that you repeat the STRUCTURE runs at least 10 to 20 times. Ln (K)-Ln(K-1) is just the difference of the estimate of the posterior probability of the data ( Ln P(D) in STRUCTURE output) for a given K value minus the estimate of the posterior probability of the data for the K-1 value. As far as I understand it.
Kind regards
Patrick
Thank You very Much Patrick Thorwarth and Piotr Androsiuk For valuable Suggestion and kind help in data calculation.
I recommend determining the most likely K:
Estimating the Delta K (ΔK) and the logarithmic probability of K, In P (K) = L (K) (Evanno et al., 2005), using the Structure Harvester website, where you just have to upload to the website the compressed folder of the results that gives the STRUCTURE program.
Or with the CLUMPAK website, where in addition to obtaining the most probable K, unite all the runs by K that you made.
However, it is also important to consider the biology of the species with which you work.
This implies that the highest Delta K value determines our best K.
- Badges
- Science method