Compare two sets of genetic correlations

contrastrg.r

library(data.table)

d1 <- fread('rgdifferences.csv',data.table=F)

#772 traits, so 0.05/772 = 6.476684e-05
#I like it better with all correlatios shown. I';ll grey out the NS ones!
d1$color <- "grey"
d1[which(d1$p <= 6.476684e-05 | d1$p_trauma <= 6.476684e-05),]$color <- "black" #significant in at least one


#How many are significant in at least one? Answer=299. 0.05/299 = 0.0001672241
table(d1$color)

#Note to caroline - adjusting for the number of traits that were significant in at least one 
#analysis only changes the bonferron therehsold to add two traits: Qualifications: O levels/GCSEs or equivalent. Seen a psychiatrist for nerves_ anxiety_ tension or depression
#imo its not worth doing that


#Since neuroticism is listed twice, I only report one of them for plotting, hence my count goes from 23 to 22 traits.
#I set the threshold to 0.00000809 to avoid this being plotted

#Therefore I keep on the strict threshold anyway, but what I do instead is plot the traits where 

#Zifference is trauma - ptsd

d1[which(d1$p <= 6.476684e-05   & d1$p_trauma <= 6.476684e-05  & d1$Pdifference <= 0.00000809 & d1$Zdifference < 0),]$color <- "blue"
d1[which(d1$p <= 6.476684e-05   & d1$p_trauma <= 6.476684e-05  & d1$Pdifference <= 0.00000809 & d1$Zdifference > 0 ),]$color <- "red"

#Driving slowly is a trait associated with PTSD but the direction 

pdf("rg_ptsd_v_trauma.pdf",7,7)
plot(d1$rg_trauma,d1$rg,col=d1$color,pch=19,cex.axis=1.45,cex.lab=1.45, xlim=c(-1,1),ylim=c(-1,1),xlab="rg LT", ylab="rg PTSD")
abline(0,1,lwd=2)

d1s <- subset(d1,color=="blue")
d1s2 <- subset(d1,color=="red")

#List only the top 10

text(d1s[c(1,4,7),]$rg_trauma,d1s[c(1,4,7),]$rg, labels=d1s[c(1,4,7),]$trait, cex= 1, pos=2)
text(d1s[c(2,5,8),]$rg_trauma,d1s[c(2,5,8),]$rg, labels=d1s[c(2,5,8),]$trait, cex= 1, pos=3)
text(d1s[c(3,6,9,10),]$rg_trauma,d1s[c(3,6,9,10),]$rg, labels=d1s[c(3,6,9,10),]$trait, cex= 1, pos=4)


text(d1s2$rg_trauma,d1s2$rg, labels=d1s2$trait, cex= 1, pos=4)
dev.off()


#Do with MVP vs ukbb..
d4 <- fread('lt_vs_mvp.csv',data.table=F)

d4$color <- "grey"
d4[which(d4$p_mvp <= 6.476684e-05 & d4$p_lt <= 6.476684e-05),]$color <- "black" #only significant in both
d4[which(d4$p_mvp <= 6.476684e-05 & d4$p_lt <= 6.476684e-05 & d4$pdiff <= 6.476684e-05 & d4$Zdiff > 0),]$color <- "red"
d4[which(d4$p_mvp <= 6.476684e-05 & d4$p_lt <= 6.476684e-05 & d4$pdiff <= 6.476684e-05& d4$Zdiff < 0 ),]$color <- "blue"

pdf("rg_MVPptsd_v_trauma.pdf",7,7)
plot(d4$rg_lt,d4$rg_mvp,col=d4$color,pch=19,cex.axis=1.25,cex.lab=1.45, xlim=c(-1,1),ylim=c(-1,1),xlab="rg LT", ylab="rg MVP PTSD")
abline(0,1,lwd=2)

d4s <- subset(d4,color=="blue")
d4s2 <- subset(d4,color=="red")


text(d4s$rg_lt,d4s$rg_mvp, labels=d4s$trait2, cex= 1, pos=2)
text(d4s2$rg_lt,d4s2$rg_mvp, labels=d4s2$trait2, cex= 1, pos=4)
dev.off()





#MVP and PGC PTSD have massive overlap 90% correlated

d2 <- fread('MVP.EUR.TOT.txt.gz_f2.premunge2.2d4b1c2c-0f1b-402b-b7dc-670725621c70.rg.results.csv',data.table=F)
names(d2)[2] <- "trait"
d3 <- merge(d4,d2,by='trait')
plot(d3$rg.x,d3$rg.y)
cor.test(d3$rg.x,d3$rg.y)