• Java, version 1.6 or later
• Dendropy, version > 4.0.0; Dendropy package which is easy to install with pip. You can find installation instructions here.
• Numpy
• Newick utilities, can be downloaded from here

1. Clone the code from this github page or download the zip file.
2. Unzip the package if you downloaded the zip file.
3. cd into the directory where QuCo code is placed.
4. cd into quart_freq and compile the code using:

make clean
make QUARTETS=1

The input required to run QuCo, is a set of posterior trees that are explored through MCMC sampling, for all the genes. If you have run multiple chains, you need to combine them into one file for each gene, and don't forget to remove trees explored during burn-ins.

In one file, provide the list of paths to the posterior trees for each gene, and this is the first input to run Quco. You can now run Quco with the following command:

./run_quco.sh [paths_to_pp_trees] [output_folder]

Each of the posterior files correspond to one gene, and it usually contains several trees (up to thousands). For example, if you run MrBayes to obtain posterior trees, you can create these files from XXX.runX.t output of MrBayes, by removing the trees corresponding to the burn-in period. MrBayes has defualt burn-in of 25%, in that case, you should remove the first quarter of the trees from each XXX.runX.t file and then combine them.

• If you have a few species and most of the trees are the same, it is not needed to save all the trees in a file. You can also pass trees with number of times observed as input to the script. In this case each line of each posterior files include first frequency, and then the tree in Newick format. Take a look at data/01/seq600/000.nex.combined.

N1 ((A,B),((C,D),E);
N2 ((A,C),((B,D),E);
N3 ((A,D),((B,E),C);
...

• Make sure that the write permissions for the output folder are correct.

The output folder contains the quartet frequencies for each gene and also inferred quartets with branch lengths and likelihoods.

The quco-out.tre has the following format, if it was needed for some analysis:

#Branch_length -best_lh best_topology -2nd_best_lh 2nd_best_topology -3rd_lh 3rd_topology
1.0949076826832758 168.9191119151238 ((A,B),(C,D)); 219.72646034989606 ((A,C),(B,D)); 219.72646185380768 ((A,D),(B,C));
1.0064805441823255 175.00172401706408 ((A,B),(C,E)); 219.72646032774875 ((A,C),(B,E)); 219.72646377537126 ((A,E),(B,C));
0.9685868730517407 178.2975624082731 ((A,B),(C,F)); 219.72646188616702 ((A,F),(B,C)); 219.72646252857652 ((A,C),(B,F));
.
.
.

Astral tree run on best quartets can be found in the output directory as quco-out.tre.astral.out . It is in newick format, as well as all other output trees.

Datasets used in the paper can be accessed via this Drive link.

quart_bin code has been adopted from this link, and the following dissertation

• Johansen, Jens, and Morten Kragelund Holt. "Computing triplet and quartet distances." (2013).