The Structure Harvester had been experiencing issues when trying to make clumpp input files for certain types of Structure analyses.
The Harvester core was updated to be more platform agnostic and to make use of unittests.
The Structure Harvester had been offline for more than a month due to a server crash.
The plotting code was pulled into its own module to allow for future extensions / changes.
The Structure Harvester paper has appeared in print, the citation has now been finalized:
Earl, Dent A. and vonHoldt, Bridgett M. (2012) STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources vol. 4 (2) pp. 359-361 doi: 10.1007/s12686-011-9548-7
I've changed the css on the website to make use of Bootstrap, from Twitter. The functionality has remained the same.
This website and the stand-alone version of the harvester are now importing the same module, harvesterCore.py, instead of simply having a bunch of copy-pasta between the two. The stand-alone version and the API are now also available on github.
The harvester will now recognize unexpected values for Est Ln Prob of Data, Mean value of ln likelihood and variance of ln likelihood. Specifically there are checks for these values being nan or inf. These errors are now reported to the user. Should this happen to your data my humble recommendation is to discard the problematic run and re-run. Thanks to An Broeck for bring this to my attention!
We now have a proper citation for you:
Earl, Dent A. and vonHoldt, Bridgett M. (2012) STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources vol. 4 (2) pp. 359-361 doi: 10.1007/s12686-011-9548-7
Updated to reflect the final publication in print (date changed to 2012).
Minor tweak that slightly speeds up processing files that contain many individuals or loci.
There are tests performed to ensure that your data conform to certain requirements before the Evanno method is applied. One of these tests checks to see if for all K, the standard deviation of the ln prob(data) values is not zero. It turns out that this test was too strict, the first K and last K may have standard deviations of zero without causing a divide by zero error (see the faq for details on the delta K calculation). Thanks to Gabriel EM Ferreira for bringing this to my attention!
Added eps and pdf formats for figure outputs. The Harvester now creates publication quality figures for you! Pdf and eps images are automatically included in the downloadable archive, and links to the individual files are provided under the images on the website.
Harvester has slightly more specific regular expression for targeting result file information.
Harvester now generates Clumpp pop files when possible (thanks to Dr. Tomas Hrbek for the suggestion).
Delta K calculation has slightly changed, see the FAQ.
The Harvester is now 100% Python.
Plots are now generated using matplotlib.
Website is slightly more standards compliant.
The Harvester now does a better job of failing if the user supplies a small number of replicates and those replicates end up having a within k standard deviation of 0. The Evanno method will not be applied in these circumstances, but the Harvester will provide a link to a 'details' page that will describe the reason for not applying the method. Hat tip to Dr. Laurence Clarke for bringing this to my attention.
The Harvester now automatically compresses all of your results into a single .tar.gz file so that you can easily download everything in a single click. Thanks to Dr. Peter de Knijff for requesting this feature.
It seems newer versions of STRUCTURE can output 90% probability intervals in the inferred ancestry table. The CLUMPP .indfile routine was not removing this data from the .indfiles and this was leading to errors in CLUMPP. Thanks to Shazia Mahamdallie for pointing this out!
Yesterday's ``fix'' in the CLUMPP indfile generation turned out to be less correct than I thought. Another hat tip to Dr. Segelbacher.
Added a step in pre-processing to convert files from DOS line endings to UNIX line endings. Users shouldn't notice any changes. Fixed a slight error in my regex for CLUMPP infile generation that would cause K=1 indfiles to be slightly malformed -- Thanks to Dr. Gernot Segelbacher for pointing this out!
I have updated the offline version of the harvester to perform the Evanno method, make CLUMPP infiles, and create gnuplot infiles. You can get it here.
We migrated the server from Fedora to Centos 5.3. Everything should be back online, please email me if you run into any problems.
I reworked the way the upload and server processing redirection occurs, if you are redirected before the Results page is finished generating, you will hopefull land at a self-refreshing limbo page, instead of being dumped at a 404 error. I changed this after an email from Dr. Alan Meerow (thanks!). Also, I think I finally fixed that zip archive directory recursion problem.
Moved to a new home on taylor0!, switched gnuplot font, made backend more idiom-compliant.
Now Featuring Automatic Evanno (2005) processing.
Maintenance Update. Fixed Typos. Redid example data. Added version numbers. Hope to add automatic Evanno (2005) soon.
Now Featuring Automatic LnP(D) vs. K Plotting with means and stdev!
Now Featuring Automatic CLUMPP .indfile generation!