Questions About Prottest

Question

[Dejian]

Hi, I have some questions about the ProtTest options (version: prottest-3.0 under Linux). I want to choose the best model for a ML tree. My command line is as follows:

java -jar prottest-3.0.jar -i msa.phy -o msa.phy.out -all-matrices -all-distributions -F -S 2 -t1 &

However, three of the ProTest options, i.e. "Tree", "StrategyMode", and "Display consensus tree" (see below), in the result file confused me. My questions are as follows:

1. What is the relationship of "Tree" and "StrategyMode"? It confuses me because I don't know whether the best model produced here is for NJ or ML.
2. When to set the "Tree" option? I think the "Tree" option is set via the "-t tree_filename" parameter which requires a tree file. If I want to find the best model for ML tree, should I provide a tree via "-t"?
3. I did not set the option of "-tc consensus_threshold". However, there is a tree at the end of the result file since I set "-t1". I wonder whether that is the "consensus tree" of bootstrapping (If yes, how many bootstraps by default? And what is the default consensus_threshold?). Additionally, does the tree mean that there is no need for further calculation in PhyML?

ProtTest options

-
Alignment file........... : msa.phy
Tree..................... : BioNJ
StrategyMode............. : Maximum Likelihood tree
Candidate models......... :
Matrices............... : JTT LG DCMut MtREV MtMam MtArt Dayhoff WAG RtREV CpREV Blosum62 VT HIVb HIVw FLU
Distributions.......... : +G +I +I+G Uniform
Observed frequencies... : true

Statistical framework
Sort models according to....: AIC
Sample size.................: 350.0 (not calculated yet)
  sampleSizeMode............: Total number of characters (aligment length)
Other options:
Display best tree in ASCII..: false
Display best tree in Newick.: true
Display consensus tree......: false
Verbose.....................: false

Answer 1

[Whetting]

I usually use the GUI version of prottest, but the parameters should not change

1) the BioNJ tree is used as a guide tree. If you want to provide a tree you can do so with the -t option.

2)Providing a better start tree is probably only beneficial if you use the fast mode (fixed tree; optimization of model parameters and branch lengths; this is the strategy used by Modeltest). Under the slow mode (optimization of model, branches and topology of the tree) you calculate everything anyway.
3) I think this tree is the starting tree. You definitely need to use another program to calculate the tree with the selected model