Modeltest on the cluster
The scripts for running Modeltest (version 3.6) and MrModeltest (version 2.2) are John Nylander's (Modelfit.pl and Mrmodelfit.pl). They have been modified so that we can use them on the cluster. In John's words, these are "Simple Perl scripts for running Modeltest or MrModeltest and PAUP by simply typing : modelfit.pl INFILE. The script will read the infile (Nexus format), launch PAUP and then Modeltest and print the selected model(s) to the screen. The output from Modeltest is printed to a separate file."
To use the scripts on the cluster:
- Create or obtain a nexus formatted data set. Make sure that your nexus formatted data set has UNIX line breaks (use TextWrangler).
- Obtain a Modelblock file. The default file from Modeltest works fine, but make sure the file has UNIX line breaks. You can download a file with UNIX line breaks here (right click and download).
- Log onto the cluster. In your home directory on the cluster, make a directory called "Modeltest".
- sftp your Modelblock file into your Modeltest folder. It is important that your Modelblock file be named "modelblock". You can use Fugu to sftp.
- sftp your nexus data set into the Modeltest folder.
- Change into your Modeltest directory. At the prompt type "runjobs.mfit filename", where filename is the name of your nexus file. The script will begin running.
- When the script finishes, an email will be sent to your UMSL email account (i.e., email@example.com). Depending on the size of your data set it may take some time to run. Your results will be printed to a file called mfit.out.
To run Modeltest with multiple datasets simultaneously, create a subdirectory within the Modeltest folder for each dataset. You can then run the "runjobs.mfit" script for each dataset from within its respective folder.
To run MrModeltest, substitute the above occurrence of runjobs.mfit with runjobs.mrmfit. You will also need to place the MrModelfit block file (name it MrModelblock) into a folder in your home directory named MrModeltest.