Post date: Dec 24, 2014 1:39:5 AM
The first tests were mostly encouraging. Accuracy was lower for the 1000 generation simulations, but for a population of 500 individuals this results in a bunch of drift! However, the initial simulations did not include genetic drift for the allele frequencies (which is important, and needs to be in there). So, I simulated additional genetic data (from the same ancestry data) with F = e(-T) (e(T) - 1), where T = no. gens / N (normally 2N but the population sizes for each ancestry type are roughly half the total). Note, for the 1000 gen. simulations, F ~ 0.8 (many fixed SNPs in the hybrids that were segregating in the parents). The script to simulate genetic data with drift from the ancestry data is sims/simGenotypesFst.pl.
As an initial test, I am running the MCMC analysis (one chain each) for the first two replicates under each demographic model (2 reps * 3 no. gens) with scale parameters of 0.1, 1, and 10 (18 MCMC analyses total). Here is one of the examples:
popanc -o outest_10F_r2f0.3gen_demog_gens10.hdf5 -m 50000 -b 10000 -t 5 -f 1 -w 0 -v 0 -c 10 /labs/evolution/projects/popanc_sims/sims/genoP0F_r2f0.3gen_demog_gens10.txt /labs/evolution/projects/popanc_sims/sims/genoP1F_r2f0.3gen_demog_gens10.txt /labs/evolution/projects/popanc_sims/sims/genoAdmxF_r2f0.3gen_demog_gens10.txt > /labs/evolution/projects/popanc_sims/mcmc/log_estest_10F_r2f0.3gen_demog_gens10
scp outest_10F_r2f0.3gen_demog_gens10.hdf5 /labs/evolution/projects/popanc_sims/mcmc/