I am studying the selection pressures of a plant pathogen in a single randomized block design using microsatellites. and therefore migration rate between the sub-populations within this trial (between which i wish to calculate differentiation) should be at a maximum, and is not a factor. for this reason i am unsure whether to use Fst or Rst, especially given that in practice when i calculate both they give me the same result (using ARLEQUIN). I am guessing probably Fst due to its lower variance, but i can't seem to find any literature on the use of population differentiation in terms of differential selection (rather than population isolation and gene flow).
Also from my reading i can see that interpretation of Fst (and its analogues) becomes problematic as heterozygosity of the markers used increases, and the maximum value Fst can reach decreases. this worries me as i am using 16 microsatellite loci, is there any way i can overcome this? is there software that can perhaps calculate standardized Gst?
Lastly i am unsure whether to use "compute distance matrix" or "conventional F-stats from haplotype frequency" when analysing my results in ARLEQUIN, these two different methods give different results and i cannot seem to find any information on which would be most appropriate for my situation.
Hi Thomas
Let me start by your last question: if you want to take advantage of the differences in the microsat allelic sizes (Rst), you should compute the distance matrix among the different alleles (i.e., if alleles 1, 2, 3 and 4 contain increasing number of repeats, the distance from 1 to 2 will be shorter than to 3, shorter than to 4). Otherwise, all your different alleles will be at a distance of 1 from each other (Fst). (I assume you already know that your alleles should be given in number of repeats for using Rst)
Now, is it possible your Fst and Rst calculations give the same result because you do not compute the distance matrix but use "conventional F-stats?
Regarding the software question, you can use SMOGD software (www.ngcrawford.com/django/jost/); which will calculate both Gst and D (Jost, 2008). Or you can use the R package mmod, which calculates Gst, G''st and D.
Hope this helps
Pablo
thanks very much Pablo, this is a great help. currently my data is given in integer allele sizes (called and binned in genemapper), i will try redefining them according to repeat sequence number if that is what fits the assumptions of Rst.
Exactly, Thomas. Rst needs the number of repeats in order to define a distance metric, whilst Fst just considers any two alleles as equal or different
If you do not have the real bp from the flanking sequences (and therefore you do not know the real number of repeats), you can just assume your smallest allele has 5 repeats (let's say) and then simple arithmetics will give you the number of repeats of the other alleles.
Cheers
I see, (i should have read the arlequin manual more carefuly as it's right there on page ten). fortunately the size range of these markers are published and it will be easy enough to redefine them in bulk in excel.
Thanks again.
Hi Thomas.
In my hands Rst is unuseful and if you look into literature is used very rarely. Idiscourage to use it, though coding allele in repeats can be anyway a good idea because allele size is taken into account in other analyses.
Essentially Rst is useful if you expect that new mutations plas a role in determining your differences, but in many cases one expects differences in frequencies only, and so is better to use Fst for its lower variance, as you said.
I suggest to use Arlequin Fst, that is the Wier & Cokcerham version of this index. In this case you might want to standardizeWC Fst it following Hedrick (2005) and Meirmans and Hedrick (2011), see below,
In alternative, you might want to look, with the program Genetix, at Robertson & Hill (RH) estimate of Fst or at RH "corrige Rafauste & Bonhomme" that should be less influenced than WC Fst by heterozigosyty.
The idea behind Hedrick standardization is very simple. Hedrick worked with Gst, for which, considering that the maximum Gst is bounded by the average homozygosity of the subpopulations, he suggested to calculate this maximum Gst by simply putting different alleles in each sub-population still maintaining their observed allelic frequencies (i.e. not changing each subpop homozygosity). Thus if one subpop have allele A1 at 0.4 and A2 at 0.6 of frequency, and the second subpop has A1 at 0.1 and A2 at 0.9, Hedrick recode the alleles of the second subpop as A3 (frequency 0.1) and A4 (frequency 0.9). After this recoding, the idea is to calculate Gst from the original data and to divide it by the maximum Gst, obtained from recoded data, to obtain a value (G'st) that is expressed as percentage of the maximum (es. G'st 0.15 would mean that is the 15% of the maximum).
Though Hedrick reasoning was for Gst, later Meirmans and Hedrick (2011) showed that the same approach can be used for any Fst related measure.
The calculation is a little tricky and you can use RecodeData (http://www.bentleydrummer.nl/software/software/Other%20Software.html) a simple routine to recode the input files so that each pop has different alleles with the procedure outlined above. So one would calculate pairwise Fst (or other measures) with the original and recoded data and obtain standardization himself by dividing each value of the "original" matrix with the corresponding value of the "recoded" matrix.
PRACTICAL NOTES RecodeData works with FSTAT files. So, you will need to convert your original files used for Arlequin to FSTAT format (using CREATE, link indicated before), use RecodeData and then convert the recoded file back to Arlequin and Genetix (again CREATE works well). Again this is a bit tricky, but it takes few minutes...
Hi, my answer might be a bit late however, I think that this is important to use a test to check if it is better Rst or Fst. This is the test I used in my paper. " Rst is a pairwise population genetic distance that is analogous to Fst, but that takes into account differences in the number of repeats between microsatellite alleles (allele size). We applied the test suggested by (Hardy et al., 2003) and implemented in SPAGeDi version 1.2 (Hardy and Vekemans, 2002) to choose the most suitable estimators. This test indicates whether or not allele sizes provide more information on population differentiation. We compared the observed Rst values with the distribution of Rst obtained after 10 000 allele size permutations (pRST). Rst would be expected to be significantly higher than the mean permuted value (pRST) when the migration rate is lower than the mutation rate (Hardy et al., 2003). A non-significant result (Rst not significantly different from pRst) would suggest that the allele size is not informative for population differentiation. Significant tests on Rst values are expected if populations had diverged for a sufficiently long time and/or if populations exchanged migrants at a rate similar or inferior to the mutation rate (Hardy et al., 2003).
- Badges
- Science method