Sunday, July 6, 2008

Software Review: Structure 2.2 GUI

The structure method introduced by Pritchard et al. in 2000 has quickly become one of the most widely-used analytical methods in population genetics and phylogeography (more than 1,000 citations in seven years!). It's popularity is sure to grow as extensions are developed and the type of multi-locus genetic data it requires accumulates for non-model organisms. With the release of a cross-platform graphical front-end in 2007, users who were scared of by its somewhat cumbersome command-line interface are out of excuses. Although veteran users may prefer to stick with the command line interface, I've found the graphical front-end to be a helpful supplement. It's particularly useful for visualizing the results of structure analyses, including generation of the iconic structure diagrams that previously required Noah Rosenberg's complementary distruct package. Don't trash your copy of distruct just yet though, you're still going to need it to produce high quality images for publication; the structure GUI produces low quality JPG images and has limited options for custom labeling. OK, now go play!

If you need support for structure, you should consult it's outstanding user manual (it's included as the readme.pdf file in the doc folder of a standard structure installation), or the wonderful tutorial that Bob Thompson posted over at the Bodega Phylogenetics Wiki. I just added a foot-note to Bob's tutorial that you might find helpful for installation of the front-end on Mac OSX machines if you have little or no experience navigating UNIX file architecture.

7 comments:

Anonymous said...

We have been using structure for the past year to examine patterns of introgression among freshwater fish species. There are some hefty assumptions that come with the model, and users should familiarize themselves with the theory backing the method.

Glor said...

Which assumptions do you find most troubling? I'm concerned that the model is designed specifically to identify disjunct populations and can't distinguish among alternative models of population structure (e.g., isolation by distance versus population fragmentation). Nevertheless, it seems no more assumption laden than other methods in population genetics.

Mark Siddall said...

Assumptions or not, STRUCTURE (without the GUI) was central to our determination that various incipient species of leeches just weren't having sex. The GUI's nice, but the output from the command line is very easy to interpret.
Siddall

Glor said...

Is this work on leeches published? I just did a quick check on your web page and web of science and didn't see anything obvious.

Anonymous said...

My comment was not meant as a critique of structure. Mark, what type of data did you use in your leech analysis, and how much data (e.g. number of genes, or microsats, etc.)?

Anonymous said...

One thing I will mention is that structure has allowed us to determine that different species of darters are having limited interspecific sex. However enough to transfer mt genomes, but not enough for the introgression of autosomal alleles.

sergios-orestis kolokotronis said...

Awesome program, indeed. Nature sometimes departs from HWE and locks a high number of alleles in homozygous combinations. That along with high levels of gene flow (Fst<0.05) can make Structure see one genetic cluster. There's also the issue of unsampled (a.k.a. ghost) populations that GeneClass2 deals well with. But I'm departing from the original posting.

I believe Mark's referring to results obtained after his 2007 paper.