Structure Harvester

[griffinia]

Dent A. Earl at UC Santa Cruz has created a great Web-based tool
called STRUCTURE HARVESTER which takes as input a zip file of your
Structure output (follow instructions online), and plots likelihood
and variance against K. It can also calculate and plot the Evanno et
al. delta K method, and generate CLUMPP infiles for you. Finally it
allows you to download a txt file with the numbers used to create the
plots. The Web version of the program is found at:
http://taylor0.biology.ucla.edu/struct_harvest/

You can download the PERL scripts for local use at:
http://users.soe.ucsc.edu/~dearl/software/struct_harvest/, but it
doesn't do the plotting nor the delta K method.

Alan Meerow

[Mrksherman]

Hi,

I am worried that my STRUCTURE runs that gave me a -1.$ as the LnP(D)
is skewing my results when I incorporate them structure harvester. My
data tends to plateau so I choose to use the Evanno et al. method. I
ran STRUCTURE HARVESTER with all the data, then ran it after removing
the runs that gave me -1.$ as the lnP(D). The resulting data gave me
different patterns (my likelihood peaked) and the program the program
did give me reliable indfiles for CLUMPP, but it did not run the
Evanno et al. method because the groups no longer had runs that were
equal in the number of replicates. So I suppose my advice when using
this program is when you think you need to use the Evanno et al.
method, rerun the runs so that you have an equal number of replicates
and include those in the zip file.


Mark

[griffinia]

I suggest that you prune those reps from your results. Just be sure
that you then even out the number of runs at each k for Evanno et al.
delta k method.

Alan

[Dent]

Hi all,

I'm the author of the Harvester, (http://taylor0.biology.ucla.edu/
struct_harvest/).

In order to run the Evanno* method in the Harvester you need to have a
continuous series of K values, all with the same number of replicates.
If you skip K values (say run K=2, 3, 5) then the Evanno routine (in
the code) will not be run. If you have different numbers of replicates
(say K=2 has 10 reps, K=3 has 10, K=5 has 11) then the Evanno routine
will not be run. These restrictions come from my interpretation of the
Evanno method.

The Harvester will still pull out all of the summary data from your
results, make a nice plot for you, and create all of your CLUMPP
indfiles, regardless of the number of replicates or the sequentiality
of the K values. The issues I mentioned only affect whether or not the
Evanno routine gets run.

Pro-tip: I saw this discussed somewhere else and I thought it might be
useful to state the details how how I implemented the Evanno routine
in my code. It may seem silly to run K=1, but if you do then you can
get results from the Evanno method for K=2. The method of calculating
delta-K for any value (k) of K uses the next value of K (k+1) and the
*previous* value of K (k-1). Not (k+1) and (k+2) as one might guess
(for verification, check the Figure 2 in the Evanno paper and compare
part A to part D. From their caption:
L'(k) = L(k) - L(k-1) (1)
L''(k) = L'(k+1) - L'(k) (2)
Thus,
L''(k) = [L(k+1) - L(k)] - [L(k) - L(k-1)], substituting (1) into (2).
Note that in part A you have K in [2, 23] and in part D you have K in
[3, 22]. If you were to use (k+1) and (k+2), as I have seen advocated
elsewhere, you would not be able to get a value for delta-K for k=22).
So you might want to run STRUCTURE for at least one value of K less
than you think is reasonable and one value of K greater than you think
is reasonable.

I'd like to thank Alan for writing about the Harvester, and invite
people to email me (my preferred email is on the Harvester webpage)
questions, suggestions, or requests regarding the Harvester.

d

*Evanno et al., 2005. Detecting the number of clusters of individuals
using the software STRUCTURE: a simulation study. Molecular Ecology
14 , 2611 - 2620 <a href="http://www3.interscience.wiley.com/journal/
118706173/abstract">link</a>

[john]

Hi!

I run python 2.7 on window 7 and structure harvester program will not respond to me.

what shall i do.
thanks

> > > You can download the PERL scripts for local use at:http://users.soe.ucsc.edu/~dearl/software/struct_harvest/ , but it

[Yohannes Besufekad]

--
You received this message because you are subscribed to the Google Groups "structure-software" group.
To view this discussion on the web visit https://groups.google.com/d/msg/structure-software/-/sjox3GOxBCMJ.
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.

[Dent]

Hi John,

I'm the author, you can respond to me off-list to get help with the Harvester software. 

Regards,

d