What's new in Structure 2.3.4

Pritchard Lab, Stanford University


The 2.3.4 release (Jul 2012) is a minor release with bug fixes, mostly in the frontend version.
  • There was a major bug with the multithreaded code in the frontend version which would stop the console version from running at odd times, or even if it ran the results appeared to be lost. This has been fixed and should help with many problems reported on the structure mailing list. This problem does not occur in the console version and all results that were completed are valid, since this bug does not affect the validity of the results.
  • Permissions: Structure needs write permissions for the folders where the results are to be stored. The frontend now checks for permissions and warns the user. In general the user does not need write permissions for the folder where Structure is installed, except when using the Tree plot function (the library/neighbor/plots folder needs write permissions)
  • Long filenames: On most operating systems there is a limit to the length of a filname, 255 characters in general. Structure now checks to see if the filenames will be too long and asks the user to choose a different name.
  • Console version: The previous console version reported a wrong version number, this has been fixed.
  • Console version: Previous versions printed a value for "Ln Like" during burnin, which shouldn't be used and on some systems it overflowed. During burnin there is no value printed for "Ln Like" only "--".
The windows installer file will automatically replace an existing installation of Structure 2.3.3 with version 2.3.4. Any files from the user stored in the install directory will still remain.

Release Note for Structure V2.3.3.

The 2.3.3 release (Jan 2010) is a minor release with a few bug fixes.
  • Assertion error when using USEPOPINFO flag. This issue has been fixed.
  • Polyploid handling. When analysing polyploid data with loci with a large number of alleles and high ploidy, previous vesions of Structure would stall and return empty files. This issue came up when the rejection method of imputing genotypes failed to find a valid sample. If it fails it is replaced by a random sampling of genotypes consistent with the data.
  • Incorrect Net Nucleotide Distances. When using the PRINTNET flag and a large number of runs, Structure would report an incorrect value for Net Nucleotide Distance between populations. The values should always lie between 0 and 1. This issue has been fixed.
In addition to the bug fixes, the frontend application for Mac is not packaged as a regular Mac Application.

Release Note for Structure V2.3.2.

The 2.3.2 release (Sept 2009) has extended the plotting features to include high quality graphs of trees. The trees show a representation of the genetic distances among the K Structure clusters. The trees are computed by applying the neighbor joining algorithm to the matrix of allele-frequency divergence among clusters (net nucleotide distance).

Release Note for Structure V2.3.1.

The 2.3.1 release (April 2009) introduces new models for improving structure inference for data sets where (1) the data are not informative enough for the usual structure models to provide accurate inference, but (2) the sampling locations are correlated with population membership. In this situation, by making explicit use of sampling location information, we give structure a boost, often allowing much improved performance (Hubisz et al., 2009). We hope to release further improvements in the coming months.

Release Note for Structure V2.2

The 2.2 release (April 2007) allows the use of dominant markers such as AFLPs (commonly used in non-model organisms) and can accommodate null alleles and other data ambiguities (Falush et al 2007). There are a few bug fixes including fixing a problem that occurred occasionally on Windows machines, causing structure to crash at seemingly arbitrary values of K. We have tidied up some aspects of the front end and C kernel. We now print out an average pairwise nucleotide distance within and between clusters, instead of the Kullback-Leibler distance which was less interpretable and less numerically stable. The new "net nucleotide distance" will facilitate drawing trees to represent the distances among clusters (see eg. Falush et al 2003 [Science]). The output format for the linkage model has been changed. We hope to release further improvements in the coming months.

Release Note for V2.2.2

The 2.2.2 release is a minor release with a few bugs in previous release fixed.

  • An occasional numerical underflow in MCMC runs fixed. When this bug is triggered, the likelihood value is mistakenly displayed as "-1.$".
  • A bug in dominant marker and null allele model fixed, the bug is trigged as user uses code "-1" in data file.
  • A window batch program is added to windows frontend package, users could customize this batch script (edit in text editors) to start Java virtual machine when the original program fails to create virtual machine due to memory allocation problem.

    • Release Note for V2.2.3

  • The 2.2.3 release fixes a bug that caused numerical problems for some data sets when lambda was very small (<<1). The problem was introduced first in V2.2.1.
  • We have also added the option to set the random number seed manually, thereby giving users the ability to repeat runs.