Version 1.1.0 released
21/05/13 08:30
We're pleased to announce the release of PartitionFinder v1.1.0.
This version contains a number of major improvements over previous
versions, and a number of bug fixes and minor improvements too.
* Support for very large datasets (1000s of genes, 1000s of taxa)
* New, faster, algorithms for finding partitioning schemes (search=rcluster and search=hcluster)
* Support for rapid model selection using RAxML (--raxml option)
* Hugely improved memory efficiency
* Output of the best partitioning scheme in RAxML format (GARLI format coming soon)
* Major improvements under the hood to speed things up
* A number of bug fixes and other minor improvements
We have a paper in the works describing the details and performance of the two new clustering algorithms, but brief descriptions are given in the manual.
This version contains a number of major improvements over previous
versions, and a number of bug fixes and minor improvements too.
* Support for very large datasets (1000s of genes, 1000s of taxa)
* New, faster, algorithms for finding partitioning schemes (search=rcluster and search=hcluster)
* Support for rapid model selection using RAxML (--raxml option)
* Hugely improved memory efficiency
* Output of the best partitioning scheme in RAxML format (GARLI format coming soon)
* Major improvements under the hood to speed things up
* A number of bug fixes and other minor improvements
We have a paper in the works describing the details and performance of the two new clustering algorithms, but brief descriptions are given in the manual.
Version 1.0.1 released
18/07/12 00:04
Please check the main page for download instructions.
We’ve added an FAQ page, which describes (among other things) how to run PF on linux, and how to use it for straightforward model selection.
This version makes a number of minor improvements.
Models in PartitionFinder and how to set them up in BEAST/BEAUti
K80: in BEAUti this is “HKY” with “base frequencies” set to “All Equal”
TrNef: in BEAUti this is “TN93” with “base frequencies” set to “All Equal”
SYM: in BEAUti this is “GTR” with “base frequencies” set to “All Equal”
HKY: in BEAUti this is “HKY” with “base frequencies” set to “estimated”
TrN: in BEAUti this is “TN93” with “base frequencies” set to “estimated”
GTR: in BEAUti this is “GTR” with “base frequencies” set to “estimated”
We’ve added an FAQ page, which describes (among other things) how to run PF on linux, and how to use it for straightforward model selection.
This version makes a number of minor improvements.
- PartitionFinder now works on Linux. We don’t have a specific download, but details are provided on the FAQs page. Email me if you have any questions.
- Fixed a bug in the Python version checking
- On the command line, you can now point the program either to the folder with your alignment and .cfg file, or to the .cfg file itself
- Lots of improvements to the error reporting, hopefully it is much more helpful for the more common errors now!
- We’ve added a ‘beast’ option to “models”. This will restrict PartitionFinder to consider only those DNA models implemented in BEAST. Note that the notation in BEAST is a little bit odd in this respect, to try and help, I’ve provided a list of the common model names below, with instructions on how to implement those models in BEAST/BEAUti.
Models in PartitionFinder and how to set them up in BEAST/BEAUti
K80: in BEAUti this is “HKY” with “base frequencies” set to “All Equal”
TrNef: in BEAUti this is “TN93” with “base frequencies” set to “All Equal”
SYM: in BEAUti this is “GTR” with “base frequencies” set to “All Equal”
HKY: in BEAUti this is “HKY” with “base frequencies” set to “estimated”
TrN: in BEAUti this is “TN93” with “base frequencies” set to “estimated”
GTR: in BEAUti this is “GTR” with “base frequencies” set to “estimated”
Version 1.0.0 released
04/05/12 11:25
Please check the main page for download instructions. This version has a couple of major developments, and a few minor ones.
Major stuff
Minor stuff
Major stuff
- PartitionFinderProtein - compare partitioning schemes for amino acid alignments! PartitionFinderProtein does for amino acid alignments what ParititionFinder does for nucleotide alignments. It can analyse all of the amino acid matrices implemented in PhyML (LG, WAG, mtREV, Dayhoff, DCMut, JTT, VT, Blosum62, CpREV, RtREV, MtMam, MtArt, HIVb, HIVw) with all combinations of the usual additions (+I, +G , and +F). You can use PartitionFinderProtein to do simple model selection (just like you might use ProtTest, or ModelGenerator), but you can also use it to do combined model selection and partitioning scheme selection. Like PartitionFinder, PartitionFinderProtein is also multi-threaded, so it takes advantage of new multi-core desktop PC’s.
- Windows support - We have been working on a windows version for a while. As of version 1.0.0 Windows is supported in both PartitionFinder and PartitionFinderProtein. Please contact us if you have any questions.
Minor stuff
- Fixed an issue with AICc calculations
- ~20% speed increase by scheduling model selection jobs more sensibly
- More helpful error messages for alignments
- Check for the correct python version before starting
- The best_schemes.txt file now includes the partitioning scheme in a format that can be added straight to the partition_finder.cfg file, useful if you want to compare user schemes to a scheme chosen by the greedy algorithm or something similar
- Much more thorough testing. Hopefully fewer bugs!
- Neater underlying architecture, so it’s easier to work with the source code.