I was wondering what people’s thoughts on this topic were: how many SNPs per RAD locus should be used for phylogenetic studies?
All agree this is important, with fewer SNPs per locus being considered more ‘robust’ or ‘true’, but I can’t find a real good study looking into it. With my own data the strongest supported trees were generated when I filtered out all loci with more than two 2 SNPs (range of 2-10). This does not make them more ‘real’, but does have an impact.
iPyRAD is interesting, as its default is to allow 20 SNPs per locus with a minimum read length of 35.
So what do people think? Are their good studies on this? How are you determining the best maximum number of SNPs per locus when filtering data for your studies?
Hi James, in my experience I gain good results by using not more than 5 SNPs/Locus. It also depends on the level you are working on. It might be a difference if you are working on closely related or distantly related taxa. I am working within a tribe and subgenus, resepectively, with good results using only up to 5 SNPs. A setting which allows many SNPs will increase the likelihood of including loci that might be the result of paralog alignment/ sequencing errors/ gene families and so on - despite the filtering beforehand. This will lead to "noise" and contradicting results in the calculations leading to lower BS support. But again, this might be different if working within a species on populational level. You can also try using the unlinked SNP files (one SNP per locus) to test the differences in your results. Good luck!
Hi James, in my experience I gain good results by using not more than 5 SNPs/Locus. It also depends on the level you are working on. It might be a difference if you are working on closely related or distantly related taxa. I am working within a tribe and subgenus, resepectively, with good results using only up to 5 SNPs. A setting which allows many SNPs will increase the likelihood of including loci that might be the result of paralog alignment/ sequencing errors/ gene families and so on - despite the filtering beforehand. This will lead to "noise" and contradicting results in the calculations leading to lower BS support. But again, this might be different if working within a species on populational level. You can also try using the unlinked SNP files (one SNP per locus) to test the differences in your results. Good luck!
This article provides useful hints on several parameters
http://www.nature.com/hdy/journal/v114/n5/full/hdy2014105a.html
What about this one?
https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-016-1431-9
- Badges
- Science topic
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics