tetrad

Phylogenetic species tree inference using phylogenetic invariants and quartet joining

Description

Tetrad employs the inference method originally developed in SVDquartets for inferring quartet trees using phylogenetic invariants, and joining those quartet trees into a super tree to provide a topology that is statistically consistent under the multispecies coalescent model.

Features

Tetrad offers a number of advantages over SVDquartets:

1. Easy installation (conda).
2. Simple command line tool.
3. Optimizes SNP information for each quartet set (see SNP sampling <#snp-sampling>_).
4. Bootstrap re-sampling samples both loci and SNPs (ideal for RAD data).
5. Fast: inference can be massively parallelized easily (e.g., MPI)

Installation

.. code:: bash

conda install tetrad -c eaton-lab -c conda-forge

Usage

.. code:: bash

# run tetrad on a snps data file from ipyrad
tetrad -i data.snps.hfd5 

# run on 80 cores distributed over 4 nodes on a cluster
tetrad -i data.snps.hfd5 -c 80 --MPI

# convert VCF to HDF5 and encode linkage block size.
tetrad -i data.vcf -o outdir -n data5K -l 5000

SNP sampling

To reduce the effects of linked SNPs on the results tetrad should be used with linkage information turned on (it is by default), so that only a single SNP is sampled from each locus. Tetrad samples a single SNP from each locus that is segregating within each quartet set of of samples, and repeats this sampling on each iteration (e.g., bootstrap) to maximize the amount of information used for each quartet resolution.