Dear all
I'm very grateful for the suggestions received.
Here are the replies:
Request:
Hi all,
I'm trying to use the approach suggested in SPAGeDi 1.2 (O. Hardy
and X. Vekemans 2007) to test for the existence of a significant
phylogeographic signal, using mtDNA sequence data. However there is an
issue that is not entirely clear to me: in cases when a Nst > Gst > 0
but the slope of pairwise Nst vs geographic distance is not significant,
what could be plausible explanations? Can anyone give me any suggestions
about this issue or can suggest some paper, book, that could be of help
if further questions or doubts storm me?
many thanks
Marco Carriso
-----
Answers:
Dear Marco,
there are two different issues in the problem you are describing. One is
whether there is some phylogenetic signal in your mtDNA data on top of the
differences in haplotype frequencies among populations (the Nst versus Gst
comparison), and the other whether the genetic differentiation that you
observe has a spatial component (does or not fit an isolation by distance
pattern). When you are applying the test implemented in Spagedi that
gives Nst>Gst, you are implicitly assuming an Island model of migration,
with GST measuring average genetic differentiation among populations due
to differences in haplotype frequencies, and Nst measuring average genetic
differentiation taking into account differences in haplotype frequencies
plus phylogenetic differences among haplotypes. Your observation that
Nst>Gst suggests that the set of distinct haplotypes that are co-occuring
whithin populations are on average more similar to each other than random
sets of haplotypes
at the total species level. Another way to put it, is to say that genetic
drift along cannot explain the data and thus the mutational process is
participating to genetic differentiation among populations.
When you are computing the slope of the regression of pairwise Nst
vs geographic distance, you are implicitly assuming an Isolation by
distance model (i.e. restricted dispersal), and if the slope is ot
significantly different than zero this means that this model should
be rejected. Hence in your case there is no clear spatial component in
genetic differentiation. It may be that the slope for Gst is significant
but not the one on Nst. This could happen simply because of the larger
variance of the estimates of pairwise Nst as compared to pairwise Gst.
best regards,
Xavier Vekemans
-- Xavier VEKEMANS
Laboratoire de Genetique et Evolution des Populations Vegetales
GEPV - UMR CNRS 8016 Batiment SN2 Universite des Sciences et Technologies de Lille - Lille 1
59655 Villeneuve d'Ascq Cedex FRANCE Tel: +33 3 2043 6753 - Fax: +33 3 2043 6979 http://www.univ-lille1.fr/gepv=0A=C2=A0=0A=C2=A0=0A=C2
Hi Marco,
you might obtain such pattern if you have a complex history of
colonisation by long term differentiated gene pools (Nst>Gst) that spread
over large and/or disjunct areas because you could have comparable
proportions of pairs of populations belonging to distinct gene pools
at short distance and at large distance. For an example, have a look at
http://www.pierroton.inra.fr/genetics/ComEvol/Atelier-Phylogeo/OHardy_phylogeography.pdfp.6.
Best,
Olivier
Olivier Hardy
Chercheur qualifiÃ© FNRS
Service Eco-Ethologie Evolutive - CP 160/12
UniversitÃ© Libre de Bruxelles
50 Av. F. Roosevelt
1050 Bruxelles
Belgique
Tel: +32 (0)2 650 6585
Fax: +32 (0)2 650 2445
email: ohardy@ulb.ac.be
marcocarriso@yahoo.it