‘Gene discordance analysis using ASTRAL on HiPerGator’

Really brief background

ASTRAL is a method used for inferrring species tree from an input of gene trees. The algorithm takes into account discordance among genes, by using dynamic programming. It splits the input gene tree into set of quartets and then searches for a species tree that has maximum number of each of these quartet topologies, assigning support values to each. The algorithm is really fast (takes about 2-4seconds to run) as there are predetermined partitions instead of searching for new ones. It is statistically consistent under multi-species coalescent.

ASTRAL III is the faster version that reduces the run time as compared to the older versions ASTRAL and ASTRAL II. Both of these method can still be found on Hipergator. The ASTRAL III paper compares and contrasts the accuracy with ASTRAL II, results are presented in a table. and are worth glancing over even if you are not a method developer. The results in the table indicate improved quartet scores. There is a new paper in bioRxiv on ASTRAL-PRO that deals with gene duplication and gene loss. If you want to learn more gene discordance theory start here and here

Using ASTRAL on Hipergator, interactively

Astral uses unrooted gene trees as input, and gives a unrooted species tree as output. To be specific, the input gene trees must be in newick format, can have missing taxa, and/or polytomies. After you log in to Hipergator, type

    module spider astral 

This gives the version number of Astral, currently on hipergator, the latest version is astral/5.6.2. Let’s load astral now. Note we can skip the previous step, and type “module load astral” and it will default to the latest version usually. It is always a good idea to double check and make sure the version is correct especially when reporting the version in manuscripts.

    module load astral/5.6.2

Let’s assume all of your gene trees are in a single file labeled INPUTGENETREES.tre (again note that it is in newick format). For a sample input file click here. The file has 6 gene trees in Newick format in a single file. You can concatenate your gene trees if they are in individual files with cat command. Let’s assume the extension of your gene trees is .tre, so you can use the following command”

    cat *.tre > INPUTGENETREES.tre

For the simplest ASTRAL run, type the following command

    astral -i INPUTGENETREES.tre -o OUTTREE.tre 2 > out.log

-i is flag for input file name -o is flag for output file name, user specified name 2 defines the format out.log saves the detailed information about the run into a log file instead of printing it on the screen, user specified name

Submitting ASTRAL jobs on Hipergator

Astral is written in java and installed on hipergator so for submitting a job, you have to allocate memory. Here is a sample submission script with the breakdown provided below.

echo Working directory is $SLURM_SUBMIT_DIR : tells you the path of working directory

cd $SLURM_SUBMIT_DIR: changes it to the working directory

echo There are $SLURM_CPUS_ON_NODE cores available. : prints out the number of cores available

export _JAVA_OPTIONS="-Xmx300M" : allocates 300 mb memory for the job.

If you want to increase the memory replace 300M by 1g ( for 1 gb), and 2 g for 2gb etc..

The other two lines are described as above

  module load astral

  astral -i INPUTGENETREES.tre -o OUTPUTTREE.tre 2>out.log

Astral support values

For understanding the support values in ASTRAL, click here for a note from one of the developer.

Branch lengths

Since Astral is a coalescent based method, the branch length on the output species tree are in coalescence units thereby measuring the level of discordance among the gene trees. A ML tree obtained by other methods (such as RAxML) has branch lengths in terms of substitutions per site.

Multi locus bootstrapping using ASTRAL

For more detailed notes click