::Edit:: This question has been answered, thank you!
I have been running *BEAST to estimate species divergence times, with one nuclear locus and one mitochondrial locus, using multiple individuals for each species. Posteriors look ok as far as I can tell, but a colleague has suggested that species trees for divergence times should be created using a single individual for each species. However, I cannot find any information stating whether a single individual/species or multiple individuals/species is preferable. Can someone shed some light on this issue for me?
Finally found the answer.
From Heled and Drummond 2009 (the paper where starBEAST is described)
"Multiple samples per species are necessary for a complete estimation. Even two samples per species are sufficient, given enough loci. A single sample means no coalescent events for that extant species and so no information to estimate population size. This may in turn have a detrimental effect on inferring speciation times and perhaps even species topology."
My suggestion is to use at least at least 3 different individuals from each species. Some researchers don't specify the number of sequenced individuals, but it's widely recommended to use more than one individual per species for phylogenetic studies. All the best!
I agree with Victor above.
The accuracy of the species tree estimation increases both with the number of individuals per species and with the number of loci. Also, it is important to keep in mind that in a diploid, the sequence you obtain for a nuclear locus actually consists of two copies that may be identical (homozygosity) or non-identical (heterozygosity). Even if the two copies are identical both need to be entered in the analysis -- the prevalence of homozygosity is highly informative for estimation of the effective pop. size!
Great, thank you for your responses. This is what I thought.
I just can't find anywhere in the manual, relevant articles etc that states what method is preferred. Would be great to have a reference for this. Do you know of any?
Hi Emma,
I checked the BEAST google group and this is the only answer I found on the subject (attached). The answer is "a diploid individual with 2 different alleles would satisfy the minimum requirement". As you probably have seen, you can run an analysis with only one individual/allele per species, however you may run into problems with chain mixing and convergence if your model is badly wrong (e.g. violation of the multi-species coalescent model). I also agree with the responses above, in that better intraspecific sampling will lead to better performance of *BEAST.
Cheers,
Santiago
https://groups.google.com/forum/#!searchin/beast-users/BEAST$20number$20of$20individuals$20per$20species/beast-users/F6Wij6QaMy4/_MaNsYJDrOMJ
Finally found the answer.
From Heled and Drummond 2009 (the paper where starBEAST is described)
"Multiple samples per species are necessary for a complete estimation. Even two samples per species are sufficient, given enough loci. A single sample means no coalescent events for that extant species and so no information to estimate population size. This may in turn have a detrimental effect on inferring speciation times and perhaps even species topology."
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics