Monday, September 19, 2011
Beyond the Fst-Qst comparison: Insights from the EGRU-blog
You are probably aware of the fact that there are many problems of Fst-Qst-comparisons to infer selection, and this method is also known to have weak statistical power. Essentially, this means that even if selection acts on a phenotypic trait, this method might not be able to detect it, and a finding that Qst equals Fst does not mean that selection is not operating on the trait in question. This low statistical power is a problem, because if one finds a positive result, one can always say that selection has operated on the trait of interest, but not much can be said if one finds a negative result. The trait might then be "neutral" and not subject to selection - or it might be subject to selection, but we cannot detect it with the current method.
There might be solutions and alternatives to the Fst-Qst-comparisons, however. At Juha Merilä's research group blog "EGRU-blog", he refers to a recently published theory-paper in Genetics, which seems like an interesting read. Although I have not read this paper in detail, it is a paper worth keeping in mind, and worth returning to in the future. I took the liberty to borrow the picture from Juha's blog and some of the text where he explains the main implications of their study:
"The main point here to note is that this method allows detection of signatures of selection also in the case where Fst =Qst: the selection in these cases (c,d) is inferred from the fact that population centroids tend to cluster according to selective regime (color) rather than their ancestry (shape of the symbol). It is also worth pointing out that the new method accounts for many other technical problems that have plagued traditional Fst-Qst comparisons. Read the paper and become enlightened!"
To this, I would like to add that it is unlikely that we will ever find the molecular method that can replace the vastly superior method of directly measuring natural or selection in the wild, or quantify quantitative genetic patterns that reveal the action of past selection. Observing something directly, and trying to quantify it, will certainly always reveal more about agents of selection and mechanisms, than indirect inferences like the Fst-Qst comparison.
This does of course not mean that these indirect methods should never be used. Far from it, and we have used such methods in the past in this, this and this study of ours, for example. But these methods can only be a first step, and if selection is inferred, it should only be considered as a preliminary working hypothesis, that needs to be experimentally corroborated. And moreover, some forms of selection, such as negative frequency-dependent selection, might seldom, or never be detectable when there is weak genetic differentiation between populations (precluding the prospects of finding a pattern where Qst < Fst), and in these cases, direct experimental field studies might be the only possible option. There is never an excuse to avoid going out the field or doing experiments in evolutionary biology, if it is possible.