This Github contains the code necessary to run the MSWE and MGWE analyses for analyzing edges

For questions please contact: [email protected]

Both programs are written in perl and should be relatively trivial to run, however, they rely on phyx and raxml which need to be compiled and can be found here:

RAxML: https://github.com/stamatak/standard-RAxML

https://www.ncbi.nlm.nih.gov/pubmed/24451623

Phyx: https://github.com/FePhyFoFum/phyx

https://academic.oup.com/bioinformatics/article/33/12/1886/2975328

Both programs run by typing in the program name followed by the name of the config file The vertebrate dataset is in the main folder to test it out. Each folder also contains the code and the files used from the paper/results of the analyses. The programs with examples are in the main folder. The data used for the examples is the Chiari et al. 2012 nucleotide data.

https://bmcbiol.biomedcentral.com/articles/10.1186/1741-7007-10-65

First step is to adjust the config file, for both the MSWE and MGWE the files are almost identical apart from a few available options to run with the MSWE.

1. The first settings are pxrmt,pxbp and pxrr and if it's installed to your bin you should be able to leave it as is, if not you need to specify the path all the way to the name of the program.

2. Next option is raxml, same as the phyx programs

3. Name of the outfile, the outfile will contain the info from the analysis, including parameters, likelihood calcs, AIC, conflicts identified, gene counts etc... This outfile will be printed to the main directory all is being run in.

4. For the species option, this is the edge/relationship you want to question. The species that make up the clade you are questioning should all be comma separated. So in the example the edge you are testing is the monophyly of: alligator,caiman,phrynops,caretta,chelonoidis_nigra,emys_orbicularis

5. The topologies option is asking how many topologies you want to test. The different relationships should be in your file and be the first relationships in that file. The program will analyze them in that order. If you have 3 you want to question, then put the 3 as your first three in you gene tree input file and put 3 as the topologies to question. If you want to identify conflict and see if there are dominant topologies, then run it with zero and it will try to identify your most commonly conflicting relationships.

6. The supermatrix is just the name of you supermatrix file, which must be fasta, if it's not you can use pxs2fa from the phyx package to convert it prior to running the program.

7. This is you model file, it recognizes the length of genes based on DNA, genename = start-stop

8. The Set is your tree set, so all the trees you want to test with the topologies in question at the top

9. Thread is how many threads should be used for RAxML

10. The test option will run the program on 2 genes instead of the whole dataset

11. The verbose option prints out a bit more info, such as which trees had to have which species removed because of missing data.

12. The Folder will be the name of an output folder for different temporary files the program generates (e.g. phyx logfile)

13. The secret mode can be used if you have run the program already and do not want it to recalculate the whole supermatrix likelihood.

14. Other modes in the MSWE should probably be turned off, this spits out pretty much all info it can for the dataset, in the case of the carnivory the site information is about 8Gb and that's only 13 taxa so it's probably better to leave it off

To run MGWE:

perl MGWE_Calc.pl ExampleMGWE.config

To run MSWE

perl MSWE_Calc.pl ExampleMSWE.config

Some Notes

1. Depending on the version of RAxML, where it's compiled, order of your sequences etc... you use the likelihood values may be a bit different, this can have a minor influenceon site counts etc... It should not influence overall results unless your likelihood Calcs are super similar.

2. If sites have identical likelihoods for the topology in question they are not counted since they do not support either tree, if the total sites counted seems lower than the total from the matrix, this could be why

3. The config file is very sensitive to being filled out correctly and it's important to check the outfile to make sure no errors have been detected.

4. Outfile and the output folder should have different names from anything in your working directory or else they will be deleted and written over

5. This repo contains the data from Walker et al. 2017 (http://www.amjbot.org/content/104/6/858.short) and Chiari et al. 2012 (link to paper above), as a result it is a pretty big repo since the exact analyses from the paper are in here to make sure everything is open.