Correlation matrix (combine 2 heatmaps using transparent method)

email titile: "heatmap of 'pheno_rg_matrix_reordered" - Carolina

Aug 7th 2020

Location on server: /space/chen-syn01/1/data/cinliu/plots/corrplot_transparent

Goal plot:

Original instructions

See attached the updated pleiotropic SNP count matrix, with antagonistic pleiotropic SNPs in the upper triangle, and agonistic in the lower triangle. I have re-ordered to be the same as Nini's heatmap plot so the labels can just be removed.

Two minor things that I could use some help updating, because for some reason I keep getting errors when I try to update (even though I'm sure they are simple things..) - Nini do you think you could help with these two things?

1) I am having trouble updating the side legend to just show integers, rather than numbers with decimal points (we are working with SNP counts, so only integers make sense here). One easy hack is just to cover up the 0.8-3.2 and 4.8-7.2 range of the legend with a white box in powerpoint.

2) We likely want different colours for the upper half vs lower half of the matrix, so for instance changing the upper triangle for antagonistic effects to red.

Original r script given from Carolina

pleitropic_counts_matrix.R

####Creating pleiotropic SNP count matrix#re-ordered to align with Nini's heatmap plot
rm(list=ls())
library(data.table)library(Matrix)library(RColorBrewer)library(ggplot2)library(gplots)
#change dirsetwd("/Users/carolinamakowski/Documents/Postdoc_UCSD/GWASgclus_UKB/GWAS_results/Mar2020_cm_fastGWAmlm/cojo_pleiotropy")
antdata<-read.table("antagonistic_pleiotropic_hits_counts_reformatted.txt", h=T, sep='\t', row.names = 1)agdata<-read.table("agonistic_pleiotropic_hits_counts_reformatted.txt", h=T, sep='\t', row.names = 1)
antdata<-as.matrix(antdata)agdata<-as.matrix(agdata)
#remove the "X" at beginning of colnames that is default in R, I don't think it likes colnames starting with a number, but rownames are okcolnames(antdata) = gsub(pattern="X",replacement="",x=colnames(antdata))colnames(agdata)=gsub(pattern="X",replacement="",x=colnames(agdata))
#make matrix symmetricmakeSymm <- function(m) { m[upper.tri(m)] <- t(m)[upper.tri(m)] return(m)}
antdata <-makeSymm(antdata)agdata<-makeSymm(agdata)
#reorder the rows/cols to be consistent with Nini's heatmap hclust resultsreorder <-c("1.motorpremotor_area","2.dorsolateralprefrontal_area","3.dorsomedialfrontal_area","4.orbitofrontal_area", "5.parsopercularis_area","6.superiortemporal_area","7.posterolateraltemporal_area","8.anteromedialtemporal_area", "9.inferiorparietal_area","10.superiorparietal_area","11.precuneus_area","12.occipital_area","1.motor_premotor_SMA_thickness", "2.superiorparietal_thickness","3.inferiorparietal_thickness","5.occipital_thickness","4.perisylvian_thickness","8.temporalpole_thickness", "9.medialtemporal_thickness" ,"10.middletemporal_thickness" , "6.ventromedialoccipital_thickness", "11.dorsolateralprefrontal_thickness","7.ventralfrontal_thickness","12.medialprefrontal_thickness")
antdata_reorder <- antdata[reorder,reorder]agdata_reorder<-agdata[reorder,reorder]
#combining the two matrices into one#lowertri will be agonistic correlation, upper tri will be antagonistica<-tril(agdata_reorder,-1) #lower triangular matrix, omit diagonals--> note this actually becomes upper part in figureb<-triu(antdata_reorder,1) #upper triangular matrix, this actually becomes LOWER part in figurec<-a+b#c<-as.data.frame(as.matrix(c))c<-as.matrix(c)

png(paste0("antupper_aglower_pleiotropic_hits_withcounts_reordered.png"),units="in", width=5, height=5, res=300)corrplot::corrplot(c,is.corr=FALSE,method="color", tl.col ="black",lim.segment=c(0,1),diag=FALSE,addCoef.col = "black",number.cex=0.4,tl.cex=0.5, cl.cex=0.5)dev.off()

pleiotropic_counts_matrix_updated_heatmaps.R

####Creating pleiotropic SNP count matrix#re-ordered to align with Nini's heatmap plot
rm(list=ls())
library(data.table)library(Matrix)library(RColorBrewer)library(ggplot2)library(gplots)
#change dirsetwd("/Users/niniliu/Desktop/2020lab/ch-email/heatmap/carolina edit/")
antdata<-read.table("antagonistic_pleiotropic_hits_counts_reformatted.txt", h=T, sep='\t', row.names = 1)agdata<-read.table("agonistic_pleiotropic_hits_counts_reformatted.txt", h=T, sep='\t', row.names = 1)
antdata<-as.matrix(antdata)agdata<-as.matrix(agdata)
#remove the "X" at beginning of colnames that is default in R, I don't think it likes colnames starting with a number, but rownames are okcolnames(antdata) = gsub(pattern="X",replacement="",x=colnames(antdata))colnames(agdata)=gsub(pattern="X",replacement="",x=colnames(agdata))
#make matrix symmetricmakeSymm <- function(m) { m[upper.tri(m)] <- t(m)[upper.tri(m)] return(m)}
antdata <-makeSymm(antdata)agdata<-makeSymm(agdata)
#reorder the rows/cols to be consistent with Nini's heatmap hclust resultsreorder <-c("1.motorpremotor_area","2.dorsolateralprefrontal_area","3.dorsomedialfrontal_area","4.orbitofrontal_area", "5.parsopercularis_area","6.superiortemporal_area","7.posterolateraltemporal_area","8.anteromedialtemporal_area", "9.inferiorparietal_area","10.superiorparietal_area","11.precuneus_area","12.occipital_area","1.motor_premotor_SMA_thickness", "2.superiorparietal_thickness","3.inferiorparietal_thickness","5.occipital_thickness","4.perisylvian_thickness","8.temporalpole_thickness", "9.medialtemporal_thickness" ,"10.middletemporal_thickness" , "6.ventromedialoccipital_thickness", "11.dorsolateralprefrontal_thickness","7.ventralfrontal_thickness","12.medialprefrontal_thickness")
antdata_reorder <- antdata[reorder,reorder]agdata_reorder<-agdata[reorder,reorder]
#combining the two matrices into one#lowertri will be agonistic correlation, upper tri will be antagonistica<-tril(agdata_reorder,-1) #lower triangular matrix, omit diagonals--> note this actually becomes upper part in figureb<-triu(antdata_reorder,1) #upper triangular matrix, this actually becomes LOWER part in figurec<-a+b#c<-as.data.frame(as.matrix(c))c<-as.matrix(c)
#original plot (I kept this for refernece)png(paste0("antupper_aglower_pleiotropic_hits_withcounts_reordered.png"),units="in", width=5, height=5, res=300)corrplot::corrplot(c,is.corr=FALSE,method="color", tl.col ="black",lim.segment=c(0,1),diag=FALSE,addCoef.col = "black",number.cex=0.4,tl.cex=0.5, cl.cex=0.5)dev.off()

##blue plotcol1<-colorRampPalette(brewer.pal(9,"Blues"))(200)
png(paste0("blueheatmap.png"),width = 13*300, height = 13*300, res = 300, pointsize = 10, font=2)heatmap.2(c,dendrogram ='none', trace = "none",density.info = "none", Colv = NA, Rowv = NA, labRow = FALSE,labCol = FALSE, breaks=seq(0, 8, length.out=200.1), key.title="",key.xlab="", key.par=list(mar=c(4,0,1,3), cex=1.0, cex.lab=1.0, cex.axis=2.0), col= col1, lmat=rbind(c(3,4), c(2,1)), #specifying the bar to be on the right lhei=c(0.1,1), lwid=c(0.1,1), notecol="black", cellnote = round(c, 2), notecex = 2, font=2)dev.off()### Red plot
col1<-colorRampPalette(brewer.pal(9,"Reds"))(200)png(paste0("redheatmap.png"),width = 13*300, height = 13*300, res = 300, pointsize = 10, font=2)heatmap.2(c,dendrogram ='none', trace = "none",density.info = "none", Colv = NA, Rowv = NA, labRow = FALSE,labCol = FALSE, breaks=seq(0, 8, length.out=200.1), key.title="",key.xlab="", key.par=list(mar=c(4,0,1,3), cex=1.0, cex.lab=1.0, cex.axis=2.0), col= col1, lmat=rbind(c(3,4), c(2,1)), #specifying the bar to be on the right lhei=c(0.1,1), lwid=c(0.1,1), notecol="black", cellnote = round(c, 2), notecex = 2, font=2)dev.off()

Combine the plots in combinedheat.pptx

Using power point and online tools, we assemble the plot. Instructions can be found in the power point slide.

use power point and use the shape function to make white triangles to over the part of the maps you don't need. Do this for the 2 plots you want to combine.
export the plots png from
Using https://onlinepngtools.com/create-transparent-png transform the background into transparent (in the site I used 8% similarity setting). Download the images and then copy and paste it here (just overlap the 2 images)!