gubbins

When we look at the pairwise FST between groups, it seems like there are regions of a few chromosomes that are recombinant between CP3 and CP4. But want to test this formally. A reviewer for Hajime's acineto paper recommended that we look at BRATNextGen or Gubbins for analysis of recombination. Both seem to be made by roughly the same group of people (at Sanger), and Gubbins is newer with supposed improvements, so that's what we'll try for. It makes a maximum-likelihood reconstruction after accounting for areas of recombination. It's a python script and was a beast to install, but Nick H was down here and helped me install it on my local machine.

I decided to analyze chromosome-by-chromosome. I wrote a script (multivcf2fasta.sh) that takes a multiVCF file and converts it to a multiFASTA file.

First I considered each chromosome for all 80 samples together, and ran it through grubbins on my local machine. For example:

run_gubbins.py chr14.intervals.fa
gubbins_drawer.py -o chr14.pdf -t chr14.intervals.final_tree.tre chr14.intervals.recombination_predictions.embl

The run_gubbins.py ran surprisingly fast, (minutes, even for the largest chromosome) compared to >24hrs for acinetobacter. Looking at the ML phylogeny + recombination blocks, the results didn't make much sense... i.e. there were just a few recombination block scattered here and there, and they didn't seem to add up to the areas I thought were recombinant based on the pairwise FST.

So, need to consider just two groups together (eg. just CP3 and CP4 samples together, without CP1 and CP2), as maybe the extra groups are throwing it all off. So (in the dadi/data/pf folder) make a CP3-CP4 VCF file and use that to make a new set of chromosomes, that just have CP3 and CP4 individuals.

Even just looking at CP3 and CP4 together doesn't give what I would have expected - I'm not seeing large predicted recombination events like I thought I would. Perhaps I don't fully understand how it works.