Post date: Jun 01, 2016 10:26:18 PM
As a quick test, I ran the following three sets of runs (2X each) through popanc (CRP is Corey Peak, not Carson Pass):
popanc -o outMTR1.hdf5 -n 5 -d 15 -m 20000 -b 10000 -t 10 -s 1 -q 1 -z 0 ../popfreqs/pntest_lychybAutos_YBG.txt ../popfreqs/pntest_lychybAutos_GVL.txt ../popfreqs/pntest_lychybAutos_MTR.txt &
popanc -o outMTR2.hdf5 -n 5 -d 15 -m 20000 -b 10000 -t 10 -s 1 -q 1 -z 0 ../popfreqs/pntest_lychybAutos_YBG.txt ../popfreqs/pntest_lychybAutos_GVL.txt ../popfreqs/pntest_lychybAutos_MTR.txt &
popanc -o outCRP1.hdf5 -n 5 -d 15 -m 20000 -b 10000 -t 10 -s 1 -q 1 -z 0 ../popfreqs/pntest_lychybAutos_YBG.txt ../popfreqs/pntest_lychybAutos_GVL.txt ../popfreqs/pntest_lychybAutos_CRP.txt &
popanc -o outCRP2.hdf5 -n 5 -d 15 -m 20000 -b 10000 -t 10 -s 1 -q 1 -z 0 ../popfreqs/pntest_lychybAutos_YBG.txt ../popfreqs/pntest_lychybAutos_GVL.txt ../popfreqs/pntest_lychybAutos_CRP.txt &
popanc -o outCLH1.hdf5 -n 5 -d 15 -m 20000 -b 10000 -t 10 -s 1 -q 1 -z 0 ../popfreqs/pntest_lychybAutos_YBG.txt ../popfreqs/pntest_lychybAutos_GVL.txt ../popfreqs/pntest_lychybAutos_CLH.txt &
popanc -o outCLH2.hdf5 -n 5 -d 15 -m 20000 -b 10000 -t 10 -s 1 -q 1 -z 0 ../popfreqs/pntest_lychybAutos_YBG.txt ../popfreqs/pntest_lychybAutos_GVL.txt ../popfreqs/pntest_lychybAutos_CLH.txt &
Results will be here: /uufs/chpc.utah.edu/common/home/u6000989/projects/lyc_hybanc/ancfreqs/. It looks like these take an hour or so to finish.
In the end, I will want to use combinations of parental populations, not single populations. Thus, I made the following combined population files (could change some of this based on entropy or PCA output):
perl combinePops.pl lychybAutos_MelWest.txt pntest_lychybAutos_BHP.txt pntest_lychybAutos_GVL.txt pntest_lychybAutos_REW.txt pntest_lychybAutos_SLA.txt pntest_lychybAutos_VCP.txt pntest_lychybAutos_WAL.txt
perl combinePops.pl lychybAutos_MelNorth.txt pntest_lychybAutos_DCR.txt pntest_lychybAutos_GLA.txt pntest_lychybAutos_LCA.txt pntest_lychybAutos_OCY.txt pntest_lychybAutos_SSC.txt pntest_lychybAutos_SUV.txt
perl combinePops.pl lychybAutos_Anna.txt pntest_lychybAutos_DOP.txt pntest_lychybAutos_FCR.txt pntest_lychybAutos_LKS.txt pntest_lychybAutos_YBG.txt pntest_lychybAutos_CPE.txt
perl combinePops.pl lychybAutos_Idas.txt pntest_lychybAutos_STB.txt pntest_lychybAutos_KHL.txt pntest_lychybAutos_SDC.txt pntest_lychybAutos_SYC.txt
Overall things looked fairly good, though some of the ESS and Gelman-Rubin diagnostics were crappy. I am trying longer runs with more chains and combined reference pops (30 jobs total).
Here is what I ran: perl ../scripts/wrap_qsub_slurm_popanc.pl
This includes the following pops.: my @hpops = ("MTR","CHL","SOP","TIC","CSP","BKM");
All with 5 chains (some might fail), and here is one example:
mv /scratch/general/lustre/u6000989/outBKM3.hdf5 ./
cd /uufs/chpc.utah.edu/common/home/u6000989/projects/lyc_hybanc/ancfreqs/
popanc -o /scratch/general/lustre/u6000989/outBKM4.hdf5 -n 5 -d 15 -m 50000 -b 25000 -t 20 -s 1 -q 1 -z 0 infiles/comb_lychybAutos_Anna.txt infiles/comb_lychybAutos_MelWest.txt infiles/pntest_lychybAutos_BKM.txt
mv /scratch/general/lustre/u6000989/outBKM4.hdf5 ./
Diagnostics were still a bit off, but things looked good overall. Thus, I am running the rest of the populations:
my @hpops = ("CLH","CRP","LAE","SWM","SOF","REF","EGP","STM");