I have a dataset of linked and unlinked SNPs, generate by PyRAD program. I would like to know which of these outputs I should use in phylogenetic analyzes.
Hi Juliana,
I don't see why you should reduce your dataset down to one snp per locus. In some analyses, like ancestry estimation, it makes sense to thin data to reduce LD among markers (i.e., where underlying models assume that loci are unlinked). In the case of phylogenetic analysis there is not real reason to do this. Ideally, you would also include all of the invariant sites so that branch lengths do not become artificially inflated.
As Sean said, if the model employed assumes that allele frequencies are not correlated, than you should sub-sample so as not to violate this assumption. You will need to read the paper describing the method and the manual for the program to determine which assumptions are being made, and what corrective actions can be applied to your data.
For example, SNAPP assumes that all columns in a dataset of concatenated SNPs are conditionally independent (so, no discernible LD). Other methods, like RAxML, do not impose the same assumption, however in this case you need to consider the implications of concatenating (as is done in your pyRAD outputs) multiple loci which potentially reflect different evolutionary histories, into a method which assumes a single genealogical history.
Short answer is, it depends. Which methods are you planning to apply for inferring a species tree from your data? What are the dimensions of your dataset (number of loci, individuals, SNPs, etc)?
Sorry I'm late to this one, but if you are still thinking about it, here's my $0.02. One option is to collapse your SNPs into haplotypes at each locus. This can be done using Chris Hollenbeck's rad_haplotyper software, which can be found here: https://github.com/chollenbeck/rad_haplotyper. The software requires a filtered vcf file and the original BAM files.
I'm not sure about the implications of this with regard to phylogenetic analyses, but I think this would allow you to retain the diversity represented by multiple SNPs at a locus, while avoiding the issue of linkage, by essentially converting combinations of SNPs on a contig into categorical "alleles".
- Badges
- Science topic
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics