Post date: Mar 20, 2015 11:6:19 PM
We could visualize evidence for hybridization (or the lack thereof) by computing phylogenetic trees with different chunks of the genome and then plotting these on top of each other like wad done in this paper on Heliconius. I have began playing with this. The program for making the plot is DensiTree (read more here), which I have a copy of in /labs/evolution/projects/lycaeides_genommix/phylo (its a java program). I wrote a script to grab from the maf file all of the sequences where at least N bases are present in all six genomes (only positions present in all individuals are retained). The script only keeps N bases even if more are present. It is /labs/evolution/projects/lycaeides_genommix/scripts/getAlignments.pl. I have done with with 5, 10, and 15 kb, the results are in /labs/evolution/projects/lycaeides_genommix/phylo. Another script in that directory computes NJ trees in R from the fasta files. This all seems to work, and the overall best trees and clade probabilities make sense, but the figures are noisy (too many different groupings are found). Part of this (not all probably) could be because I am using NJ trees, so I might want to try ML trees. There is now a R package for this called phangorn. This package also includes a simplified version of DensiTree as a function that might be nice to try.