Analyses with SMC++ and fastsimcoal both indicated the presence of a
strong bottleneck (most likely founder effect) in all three populations.
These bottleneck effects would have contributed to drift, which,
combined with low migration rates would have resulted in the clear
divergence observed in population structure analyses. Here we used
fastsimcoal analyses to disentangle these effects. To do this we
constructed a model (called no_bott) that was based on the best
fitting model, but, where all changes in Ne were removed, effectively
keeping the population size the same as the ancestral population with
subdivision on splits. This model is detailed in
3.out.no_bott.est and
3.out.no_bott.tpl
To work out input parameters for this model we took average Ne values from the 95% confidence interval ranges obtained from the best fitting IMc model. Then, assuming that the overall population maintains its ancestral size we apportioned Ne into IN, NO and SO populations so that their final values matched the relative sizes in the most recent epoch in IMc.
Simulations with this model were generated using
fsc2705 -t 3.out.no_bott.tpl -n 1 -e 3.out.no_bott.est -E 10 -G x -c 0 -k 10000000After converting to vcf they were used to calculate genome-wide Fst
values between all pairs with plink2. This was done separately for each
independent simulation.
plink2 --vcf ${f} --fst site --pheno phenotypes.txt --allow-extra-chr --out fst/${s2}_no_bott.fstThe boxplot below shows that these Fst values are much smaller than
those for the full model (IMc) that includes population size changes.
Thus indicating a strong contribution from the bottleneck to drift and
population divergence.
