Dear all,
I am using the linkage-disequilibrium method implemented in the NeEstimator software; to compute Ne values for several fish species all collected from the same geographical area.
Fst values across the focal species are sometimes significant but relatively low: the maximum value is a 5-6% differentiation explained by sampling sites. Therefore, I wanted to estimate a "global" (~ meta-) Ne, i.e. at the total study area scale (one value per species), just to compare these Ne value between species on the area. Molecular markers are of the same nature (110 to 150 biallelic SNPs for each species, developed based on a discovery panel from the study area) but as we might expect, gene diversity varies between species, with expected heterozygosity ranging from on average 0.25 to 0.40.
So I had two questions:
(1) eventhough sub-structure might (slightly?) bias Ne estimation at the whole geographical area scale, what would you say about the relevance of comparing Ne between different species?
(2) LD-based estimation of Ne could also be biased by highly unbalanced allele frequencies (e.g., MAF for a locus very close to 0). Have you faced such an issue; and would you have any piece of advice regarding this particular feature and the resulting interpretations?
I have already consulted the following papers; which I am enclosing here if anyone is interested:
doi: 10.1002/ece3.485 (about Ne and meta-Ne estimations)
doi: 10.1111/j.1755-0998.2007.02061.x (Waples, 2008 - LDNe)
doi:10.1098/rstb.2005.1682 (Wang, 2005)
doi: 10.1007/s10592-005-9103-8 (England, 2006)
doi: 10.1007/s10592-005-9103-8 (England et al., 2005)
I could estimate Ne per sampling site or per cluster when these exist. However, I would suffer of low sample size (typically 15-40) and when giving this a try, I usually got nothing but infinite values and/or CIs.
Thank you very much for any help! :-)
Best regards,
Chrystelle
Hi Chrystelle,
Interesting questions. Couple of things: what is the underlying population structure of these species without considering sample location? Have you performed analyses such as a PCA, clustering, or genetic distance? Estimated the number of migrants or migration rates? That way you can identify potential genetic groupings that would serve as a more appropriate units to estimate Ne. It would be more biologically-based than sample-based.
With LDNe it is often advisable to estimate Ne using different thresholds of MAF. I believe the default is usually 0.02. That will give an indication of how much MAF is affecting your estimates of Ne.
In theory Ne estimates for different species are comparable. Are the same exact loci being used across species? There are a number of studies that have pulled together Ne estimates from different species and compared them, so I would say they are comparable. The theory is the same.
Try estimating Ne at the site-level and across sites and report both with a fair discussion of what the limitations of combining site localities with slight substructuring might be. Check out some of Andrew Whiteley's paper's on brook trout: he has written extensively about this topic. Also, Wang's recent paper in Molecular Ecology suggests his sibship based estimate of Ne implemented in Colony might be more accurate and does not have the same assumptions of LD-Ne, but requires a larger sample size.
Good luck.
Thanks a lot Justin,
This confirms a part of what I was thinking. Indeed I need to try the Ne estimates on genetic clusters. Usually when a species displays genetic structure, this is either hierarchical or with more isolated headwaters (we are working on a river basin). However, I was afraid of losing a lot of statistical power, being still confounded by immigration, or not being able to compare species on a common basis if genetic clusters differ between species.
Thanks again for the paper suggestion! I am currently trying Migrate-n and will try Colony if all requirements are met.
All the best,
Chrystelle
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size