Including reference genome in SNP-based phylogeny

Question

[maxrwjones]

Hi all,

I have a set of ~15x coverage whole genome resequencing data for >200 accessions of a crop species. I have produced a SNP-based phylogeny of these accessions via the following operations:

1. bowtie2 mapping against reference genome
2. bcftools mpileup
3. bcftools call
4. vcftools to retrieve SNP-wise and accession-wise statistics, analysed in R
5. filtering using vcftools
6. vcf2phylip.py
7. IQtree to produce final phylogeny

My question is, given that I know the reference genome's state for every included SNP, it seems there is enough information to treat it as an accession in its own right. So is there a way to include the reference genome in the final tree? Seems like IQtree should have an option for this? Or maybe even bcftools or vcftools?

Many thanks,

Max

Answer 1

[Michael]

Hi,

That should be possible but somehow the sequence of the reference needs to included in the input given to IQ-Tree. It sort of is there already in the VCF via the REF allele. If you add a "dummy" sample to the vcf with genotype 0/0 or 0 (depending on ploidy of your file) that should give the right input to vcf2phylip to achieve it. Also, when selecting a model, make sure it has "+ASC" in it (Ascertainment correction), otherwise branch lengths might be overestimated.

Greetings to Norwich...

[maxrwjones]

Hi Michael,

Yes exactly! I wondered if perhaps IQ-tree could do this 'under-the-hood' but if not then adding a dummy sample with all loci set to 0/0 is the next best thing - I'll give it go. Do you think I should set the read depths and qualities to arbitrary numbers that pass my vcftools filters? Otherwise the whole dummy individual would get filtered out.

Yep, I have +ASC enabled already, thanks you for the heads up.

Greetings to you too - in Norway I see?

[Michael]

Do you think I should set the read depths and qualities to arbitrary numbers that pass my vcftools filters?

Good point! Yes I think so. Once you have a sequence alignment, the VCF statistics do not matter any more.

Yes, Greetings from Bergen, Norway. Just remembered visiting JIC and Norwich a while back (that was about 15 years at least I think).

Cheers Michael

[maxrwjones]

Excellent, I shall try and figure out a little script that appends a dummy sample to each line (i.e. each SNP).

Ah lovely! I moved here ~3 years ago for my PhD and am really enjoying it, it's a great institute to work at.

Cheers, Max

[maxrwjones]

I guess another alternative is to remap raw reads from the reference genome against itself... i.e. treat it like any other sample throughout the whole pipeline.