Post date: Sep 19, 2019 2:46:57 AM
Generate summaries (in output):
## posteriors for hyperparams
## AC sex, AC survival, MM sex, MM survival, RG, GB
perl calpost.pl o*ch0.hyp.txt
Estimates (PVE, PGE, n-gamma; median and 95% ETPIs
0.70459035 0.38902849 0.98897212 0.7417148 0.333193185 0.974374535 45 9 213
0.43126315 0.091165169 0.85202212 0.3838379 0.0020084627 0.92409651 77 1 267
0.73037215 0.439722175 0.992582005 0.6725727 0.332746275 0.958562155 30 8 109
0.26291945 0.041003074 0.7922405 0.63618645 0.0459106305 0.96830823 7 1 67
0.4800296 0.27012309 0.714918335 0.3710514 0.1915863 0.756545735 6 1 32
0.3039092 0.141404245 0.507983325 0.5086195 0.13365228 0.930705625 20 2 184
## pips
perl grabPips.pl o_mod_g_tchum_AC_ph1_ch0.param.txt
perl grabPips.pl o_mod_g_tchum_AC_ph2_ch0.param.txt
perl grabPips.pl o_mod_g_tchum_MM_ph1_ch0.param.txt
perl grabPips.pl o_mod_g_tchum_MM_ph2_ch0.param.txt
perl grabPips.pl o_mod_g_tchum_ph1_ch0.param.txt
perl grabPips.pl o_mod_g_tchum_ph2_ch0.param.txt
Plots and summaries in R:
## plot pips across traits
pf<-list.files(pattern="pip")
pips<-matrix(NA,nrow=11586,ncol=6)
for(i in 1:6){
pips[,i]<-scan(pf[i])
}
tits<-c("AC, Sex","AC, Survival","MM, Sex","MM, Survival","Color, RG","Color, GB")
scaf<-scan("scafs.txt")
cs<-rep("gray",11586)
cs[scaf==7748]<-"red"
pdf("tchumPips.pdf",width=8,height=6)
par(mfrow=c(2,2))
par(mar=c(4,4,2,1))
for(i in c(2,4:6)){
plot(pips[,i],col=cs,pch=19,ylim=c(0,1),xlab="SNP",ylab="PIP",cex.lab=1.3)
title(main=tits[i])
}
dev.off()
My interpretation:
First, overall PVEs (color was mapped with all individuals, not by treatment):
Survival AC = 0.43 (0.09-0.85)
Survival MM = 0.26 (0.04-0.79)
Color RG = 0.48 (0.27-0.71)
Color GB = 0.30 (0.14-0.51)
So, higher PVE for AC then MM, and in the basic null model it fits first based on the kinship matrix, the difference is even more striking (uncertainty has likely shrunk the difference) with PVE of 0.34 for AC vs <0.01 for MM (adding SNPs with measurable effects makes both go up and gives the results from above).
The PIP plots are perhaps a bit less compelling (see attached). LG8 is colored red, the rest of the genome is in gray and all are scaled for 0 to 1. There is some color signal on LG 8, and survival too, but survival doesn't smack you in the fact. However, I suspect this is in part because the real loci aren't there and the signal is being spread out (melanic genome will fix this). A few summaries support this.
Number of QTN on LG8:
Survival AC = 8.4
Survival MM = 1.3
Color RG = 0.92
Color GB = 4.1
Let me know what you think, but I would not take this too seriously (though suggestive in a positive sense) until with have the results based on the melanic genome (especially with regard to localizing the signal, the PVEs might be a bit more robust though we might simply be missing stuff in the deletion due to weak LD).