Inferring bottlenecks using IMa2?
86 views
Skip to first unread message
JamesH
May 15, 2012, 9:36:32 PM5/15/12
to Isolation with Migration
Dear Dr Hey and group,
I have microsatellite data for many populations of an insect that
occurs in the arid interior of Australia and also in the coastal
cropping regions. Frequency - based analyses (bottleneck and BayesAss)
indicate that some populations have undergone recent bottlenecks, and
also that recent migration is likely to have occurred between inland
and coastal populations. This makes sense in terms of the biology - in
the desert they rely on ephemeral host-plant resources that flourish
after rain and then die off. It has been suggested, however, that the
allele frequency signals from both events could be similar. I'm
therefore trying to use IMa2 as a coalescent approach to try and tease
apart these two processes. I can only use two populations at a time
(the program crashes if I use three, and I don't have that many loci
anyway), I have been selecting two populations with relatively high
pair-wise Fst's whereby one has a signal of bottleneck, and both have
inferred recent migration. So far I have been coding the tree as
(pop1,pop2):popAncestral. In exploratory (short-ish) runs I tend to
get small values of q for both the derived populations, and a large
value for the ancestral, with relatively high estimates of m. I
realise that IMa2 allows populations to change in size through the
MCMC chains but my questions are:
1. Is it possible to infer bottlenecks using the program?
2. If so how? Do you need to code further periods since divergence
into the tree? or can you infer based on the posterior distribution of
q values of the diverged populations?
Perhaps I'm trying to do something that the method does not allow for,
but I thought that I should seek expert opinion before coming to that
conclusion!
Many thanks,
James
I have microsatellite data for many populations of an insect that
occurs in the arid interior of Australia and also in the coastal
cropping regions. Frequency - based analyses (bottleneck and BayesAss)
indicate that some populations have undergone recent bottlenecks, and
also that recent migration is likely to have occurred between inland
and coastal populations. This makes sense in terms of the biology - in
the desert they rely on ephemeral host-plant resources that flourish
after rain and then die off. It has been suggested, however, that the
allele frequency signals from both events could be similar. I'm
therefore trying to use IMa2 as a coalescent approach to try and tease
apart these two processes. I can only use two populations at a time
(the program crashes if I use three, and I don't have that many loci
anyway), I have been selecting two populations with relatively high
pair-wise Fst's whereby one has a signal of bottleneck, and both have
inferred recent migration. So far I have been coding the tree as
(pop1,pop2):popAncestral. In exploratory (short-ish) runs I tend to
get small values of q for both the derived populations, and a large
value for the ancestral, with relatively high estimates of m. I
realise that IMa2 allows populations to change in size through the
MCMC chains but my questions are:
1. Is it possible to infer bottlenecks using the program?
2. If so how? Do you need to code further periods since divergence
into the tree? or can you infer based on the posterior distribution of
q values of the diverged populations?
Perhaps I'm trying to do something that the method does not allow for,
but I thought that I should seek expert opinion before coming to that
conclusion!
Many thanks,
James
jhey
Jun 12, 2012, 12:29:44 AM6/12/12
to Isolation with Migration
IMa2 only implements a model in which the population sizes are
constant in each time period, so it is not a great tool for studying
bottlenecks. You can try the original IM program which lets you
assess population size changes. See Hey 2005
jhey
Reply all
Reply to author
Forward
0 new messages