I have a question about the best way to test for violations of Hardy-Weinberg Equilibrium (HWE) among microsatellite loci for a species that is continuously distributed across a study area and showing IBD. We are looking at how landscape features affect gene flow among Eastern Indigo Snakes across a 25 x 50 km study area. We have 110 samples and about half are clustered in the southern half of the study area. A spatial correlogram of individual genetic distance shows that spatial autocorrelation among samples becomes non-significant at 5-10 km. We have used COLONY to identify full-sibs and found about 15 full-sib families although family size was usually two (max. four). There is significant IBD within our study area. STRUCTURE identifies K=4 with all 110 samples but when we randomly exclude all but one full-sib from each full-sib family STRUCTURE identifies K=1-2. We suspect that these STRUCTURE results are the result of neighborhood effects and IBD, respectively.
When we test for violation of HWE at our 15 loci, four have significant violations of HWE. Estimated null allele frequencies at these four loci are 6-15%. When we randomly excluded all but one member from each full-sib family, these four loci were still significantly out of HWE.
In a situation such as this, is it appropriate to test for HWE using all samples? I know that in systems with discrete populations researchers often test for HWE within each population, since violations may represent a mixture of multiple populations. But any designations of “populations” in our study area seem very arbitrary (e.g., driven by sampling intensity rather than the distribution of individuals).
Does anyone have any suggestions about the appropriate way(s) to test for HWE in a system such as ours?
Thanks,
Javan Bauder
Can you transform your data to a random distribution? Then HW will not be violated. Or increase the population size? :)
Were your samples taken randomly, or had to do with a specific distribution pattern? If the answer is yes to the second part of the question, I assume that if half of the samples are clustered in the southern half of the sampling area, then it would be too forceful to transform the data into a random distribution.
I think Javier is right. So maybe increase the population size if possible :)
Thank you for your suggestions! Our samples were collected opportunistically. Eastern Indigo Snakes are difficult to find so our cluster of points occurs around our radio telemetry study area where we spent most of our field effort (and where telemetered snakes could lead us to other snakes during the breeding season). Most samples were collected purely by chance. Our guess is that the true distribution of snakes within our study area is more-or-less continuous. Unfortunately, our field work is complete so the 110 samples are all we have to work with.
Javan, unfortunately, HWE assumptions are based on some arbitrariness in define populations limits, I'm not sure if you can do that without either the influence of demes in a whole population or subset of it. At this point, with no possibilities to increase your sample size, I think you can use your individual IBD results to find subsets and test such deviations, as apparently, you did. Also, I indicate Pedigree (MCMC test for family fullsibs) as an alternative for the Colony, just to have sure about the families you got, because I think in your case the parents are unknown, right?
In addition, it will be interesting you use BayesAss to test more likely gene flow rates among you fictional populations after to define it. Another interesting test will be to use the bootSVD R package to test the importance of such loci in your dataset throughout PCAs. Finally, what about the LD test for your loci? Were they good?
Good luck!
Interesting papers and links with a little about:
Lecocq et al. 2015 for individual IBD
BayesAss for recent migration test: http://www.rannala.org/inference-of-recent-migration/
BootSVD: https://cran.r-project.org/web/packages/bootSVD/bootSVD.pdf
Article Conservation genetics of European bees: new insights from th...
Thanks for your response, Wilson! I did test for LD between pairs of loci. Of 105 pairs, only 2 had significant LD, which seemed surprising with four of 15 loci out of HWE.
Pretty interesting Javan. At first, I would recommend you to discard those linked loci because you can't assume independent segregation between them or neutrality. In my experience, we usually got a few loci out of HWE either for continuous or discrete populations, large or small, particularly for microsatellites, because they frequently aren't 100% neutral. They don't change the results too much, but it should be used with caution. In addition, You can try to figure out what is leading those loci to deviation using tests based on neutrality assumptions, like bottlenecks, and then compare the results against the other loci. If you got different results between loci in/out of HWE, then you can take some consideration about natural selection over some neighbor region; unfortunately, I think it is like impossible to prove just with microsatellites unless you got some genomic data.
I hope you can find an answer.
Good luck
You expect HWE in a panmictic population. If there is population structure (IBD), why should you test for HWE?
Hi Miguel,
You raise an excellent point that I had not considered. I primarily wanted to test for HWE to see if my microsatellite loci were neutral (or close to neutral) markers (and because everyone does it in landscape genetics studies!). But since all landscape genetics studies are looking at some degree of population structure (e.g., IBD or isolation-by-resistance), is there really any reason to test for HWE in this context?
Thanks,
Javan
Hi all,
Miguel has raised a good point. As I said in my last comment, frequently you can get some loci in deviation from HWE because its assumptions are usually violated by substructure, inbreeding, mutation, genetic drift or even due genotyping errors. However, it's important to be tested to isolate loci and populations in which such deviations happen. In your case, I think it's important, at first to prove your loci has no bias itself; second to isolate such loci and test them for other effects as linkage disequilibrium, mainly if you are describing a new set o loci or proceeding some cross-amplification; finally, I believe that it is important to tested HWE in landscape genetics or classic population genetics because you can identify deviations from HWE assumptions for subpopulations or demes after their delimitation, then you can choose to perform specific tests or not.
That's my point of view.
I hope it can help you to take the best decision.
Best
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology