Hi, I have this simple but wired question that how many of you ever compared the results of the same data set using both reference-based and de novo approach, and get more loci/SNPs using the latter one.
In the paper J. Paris, J. Stevens, & J. Catchen. Lost in parameter space: a road map for Stacks. Methods in Ecology and Evolution, 8(10):1360-1373, 2017 They observed more loci/SNPs acquired from de novo mapping, while I had the opposite results for my Arctic charr data set using the Canadian Arctic charr genome.
I have found similar reports (more loci from reference-mapped) from phylogenetic studies (e.g. Tripp, Erin A., et al. "RAD seq dataset with 90% missing data fully resolves recent radiation of Petalidium (Acanthaceae) in the ultra‐arid deserts of Namibia." Ecology and Evolution 7.19 (2017): 7920-7936. ) and for their studies, the two approaches also provide different resolutions.
It is so obvious that people prefer to use reference-based mapping as it provides more detailed info for downstream analysis such as functional and ontology. However, I got clear pictures of the population structure using de novo but not in reference mapping. Therefore I would love to know your suggestions. Thank you very much!
Best regards,
Han Xiao
¿Did you follow the procedure to obtain the optimal set of parameters? And, in both instances (de-novo and reference-based) ¿did you evaluated the presence of outlier loci? Sometimes resolution of population structure may be affected by allelic frequencies that behave different from what is randomly expected.
Thanks for your reply and suggestions! Yeah for the de Novo we do follow the procedures but we didn't get the curve increasing with a plateau but decreasing all the time. It seems not only happens to us. Therefore we can not choose M=1 as it underestimates heterozygous. However as we decide to choose M=4 to merge paralogs, it might be the reason why reference-based method recover more SNPs.
Remember that while natural selection produces differential fixation of alleles within a population, mutation and drift tend to generate random fluctuations in allele frequencies between generations within that population. I may suggest you this paper, it very clarifying about the subject: Article Latitudinal divergence in a widespread amphibian: Contrastin...
What I was suggesting is to make tests of outlier loci (Bayescan, Arlequin, PCAdapt) to identify (and then purge from any of your datasets, denovo and referenced) the sites that me be under putative selection (not neutral processes). Genetic structure may look clearer appealing just to neutrality.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping