‘Wake… from your sleep‘ sang Radiohead. If you are working with microsatellites, it is about the time – there is a blow coming on your direction. In a recent issue of Molecular Ecology, Lou Joost lays out the mathematical details as to why Fst/Gst- estimators of genetic differentiation do not actually measure what they are envisioned to measure when genetic diversity is high. Same goes with the heterozygosity. As he points out with crisp clarity, the traditional estimators are often suggesting entirely misleading estimates of diversity and differentiation. For instance, in one particular empirical study, the between population differentiation was indicated be at 5% (Fst), while correct calculation placed the estimate to 100%.
In a way, this has been in the air for a while, but this might be the first paper where the problems are pointed out in a way that they might actually be understood also by less theoretically inclined geneticists. The particularly nice feature of this paper is it’s constructive, non-preaching tone: alternative and better behaving measures are derived and presented.
After taking home the new doctorine, one is naturally left to wonder how should we deal with those mountains of publications where the measures are wrong? Likewise, how long will it take before the new measures are deployed as a standard in empirical studies and implemented in standard software packages? Above all, I was left to wonder what this means from the perspective of Fst/Qst comparisons, and inference drawn out of them?
Oh yes, there was also Déja vu here: is it a coincidence that it seems to be ‘always’ these ratios that lead to hangovers? Namely, there are many examples in the history ecology, systematics and genetics where ratios have – after initial enthusiasm – been found to lead us into ashtray. Looks almost as if our strive towards intuitive measures is bound to backfire.
Jost L (2008) GST and its relatives do not measure differentiation. Molecular Ecology 17:4015-4026.
9 Comments
As a follow up, check this is too: Ryman, N. and Leimar, O. 2008. Effect of mutation on genetic differentiation among non-equilibrium populations. Evolution 62: 2250-2259.
A pdf can be downloaded from here: http://www.zoologi.su.se/research/leimar/fulltext/RymanLeimar2008.pdf
After reading Ryman & Leimar paper and discussing with both Nisse and Olle, I am inclined to back off from my enthuastic first reaction on the Joost-paper. According to these discussions, the Joost-solution does as little good to solve the problem as did the Hedrick-correction. A rebuttal is on it is way I have been told.
So where do we stand? Read the Ryman & Leimar paper – that is the best game in the town. For the time being.
Please say, where are we re. this issue that Jost 2008 raised. Did someone write a rebuttal? N. Crawford just advertised a software on Evoldir to calc. Jost’s measures.
Thanks in advance,
Regards,
I am not up-to-date what the status of the rebuttal ms is. I guess we just have to wait and see how this unwinds.
Ryman and Leimar’s rebuttal is in press, along with my response, and along with another study which supports my approach. Don’t fall for Ryman and Leimar’s misunderstandings, try numerical examples and see for yourself which one is right. Their “criticism” of Gst’ is also completely mistaken. Notice that they never look at actual examples of allele frequencies, nor do they analyze the mathematics of diversity. I’ll be glad to discuss this more after their article and my reply are out. Meanwhile, try some examples! Differences are greatest with high-diversity, high-differentiation systems.
You might also watch the ecology literature in this issue. I have made the same arguments there, and there is a Forum in press in the journal Ecology arguing about my claims. The mathematical issues are the same in both fields.
Best, Lou
Thanks for this Lou. Looking forward to see the papers!
By the way, there is some discussion of these issues on the Nature population genetics forum:
network.nature.com/groups/popgen/forum/topics/4220
I hope interested or opinionated readers will post their comments in that Forum.
I have informed Ryman and Leimar about the Nature forum in the hope that the discussion can be two-sided. Maybe together we can reach a general agreement about most of these issues.
One other comment, it is important to note that Gst is still useful for some purposes (such as estimating migration rate). It fails miserably as a measure of differentiation, though!
Best, Lou
Now the Ryman & Leimar rebuttal is out:
http://www3.interscience.wiley.com/journal/122352803/abstract)
as is another related paper
http://www3.interscience.wiley.com/journal/122352783/abstract
and Jost response to them
http://www3.interscience.wiley.com/journal/122352785/abstract#b2
Enjoy!
Continuation:
http://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2010.04784.x/abstract