Be notified of new releases
Create your free GitHub account today to subscribe to this repository for new releases and build software alongside 28 million developers.Sign up
- k-means is no longer available for most data types, see here for details: 19d7fe4
- Added output format for IQ-tree
- Added a few things to help along the way for a Debian package
This is the first full release of PartitionFinder version 2.
WARNING: we are aware that the k-means algorithm (i.e. search=kmeans) does not perform well on some simulated and empirical datasets. We are looking into this, but until then we strongly urge people to avoid using this algorithm to make biological inferences. Instead, partition things by codon position and locus. We will post updates on the PartitionFinder google group. We have left the kmeans algorithm available to facilitate other people who wish to research and improve upon it.
Changes since the last pre-release:
- a few speedups for analyses of very large datasets
- the default --rcluster-max is now the larger of 1000 and 10x the number of data blocks
Two small changes
- Fixed a bug that would lead some analysis to quite due to an error in RAxML. This was a failing assertion in RAxML, that Alexis fixed recently.
- Made the estimation of branch lengths on the starting tree a little more accurate.
A couple of changes:
- The manual is now completely up to date
- PF2 now includes a dependency checker. This will hopefully catch the common error where people download PF2 and just try to run it like PF1 without reading the manual. It checks for necessary dependencies, and if they don't have them it points them in the right direction (either get Anacondo distro, or RTFM).
- ML starting trees are now the default for all analyses. This is much better than using other trees, and the additional analysis time is minimal. Issues with PhyML are now sorted out.
- Confusing options w.r.t. TIGER rates are now sorted. Entropy is now the only option for DNA and protein datasets. Morphology datasets can use TIGER or Entropy. The default is the latter, and to use TIGER you specify --kmeans tiger at the command line.
- Morphology models have been checked and are all correct.
- Morphology is now supported. See the manual. Some default settings may still change, but for now you are free to experiment. Everything is described in the manual.
- Fix a number of small bugs in reporting output to users (on screen: the output in files was fine).
There is one known bug, which occurs when you try and specify an ml starting tree (with --ml-tree) but use PhyML for your model selection. This is an open issue, which I hope to get to soon...
This release reflects a couple of important changes.
- PF 1.1 is now tagged and archived. PF2.0.0 has replaced it on the master branch. So, for bleeding-edge copies, go to the master branch.
- This release includes new binaries for PhyML and RAxML, including instructions on how to compile your own if the included binaries don't work. The binaries are all in the /programs folder, and we have done our best to include versions that will work on mac, linux, and windows. The instructions for compiling your own binaries are in the same place.
- I fixed a small bug where PF2.0.0 would crash if you ran the greedy algorithm with a single partition.