plate <- function(name, panel, height=140) { for (fmt in c("pdf","png")) { path <- file.path(root,paste0(name,".",fmt)) if (fmt=="pdf") pdf(path,width=183/25.4,height=height/25.4,pointsize=8,useDingbats=FALSE) else png(path,width=183/25.4,height=height/25.4,units="in",res=300,pointsize=8,type=png_device) par(mar=c(3.8,4.9,2.1,1.05),mgp=c(2.4,.62,0),las=1,bty="l",cex.axis=.9,tcl=-.23) set.seed(45) panel() dev.off() } } plate("Figure1",function(){ layout(matrix(c(1,2,3,3),2,byrow=TRUE),heights=c(1.1,1)) x <- subset(t,accession=="GSE149507") x <- reshape(x[c("patient","tissue","CIP2A")],idvar="patient",timevar="tissue",direction="wide") stopifnot(nrow(x)==18) plot(NA,xlim=c(.7,2.3),ylim=range(x[-1]),xaxt="n",xlab="",ylab="Published log2 expression") axis(1,1:2,c("Adjacent lung","SCLC")) segments(1,x$CIP2A.normal,2,x$CIP2A.tumor,col=adjustcolor(pal[1],.45)) points(rep(1,18),x$CIP2A.normal,pch=16,col=pal[1],cex=.65) points(rep(2,18),x$CIP2A.tumor,pch=16,col=pal[1],cex=.65) ttl("A GSE149507") x <- subset(t,accession=="GSE60052") strip(list(x$CIP2A[x$tissue=="normal"],x$CIP2A[x$tissue=="tumor"]), c("Normal lung","SCLC"),"Published log2 expression") ttl("B GSE60052") par(mar=c(4.0,7.1,2.2,1.05)) y <- 2:1 plot(te$mean_difference,y,xlim=c(0,6.8),ylim=c(.58,2.55),xaxt="n",yaxt="n",pch=19,col=accent, xlab="",ylab="",bty="n") axis(1,0:3) axis(2,y,te$accession,las=1,tick=FALSE) segments(0,.67,0,2.22,col="grey55",lty=2) segments(te$ci_low,y,te$ci_high,y,lwd=1.6,col=accent) points(te$mean_difference,y,pch=19,col=accent) text(3.55,y,sprintf("%d / %d %+.2f [%.2f, %.2f]",te$n_tumor,te$n_normal,te$mean_difference,te$ci_low,te$ci_high),adj=0,cex=.88) text(3.55,2.36,"Tumor / lung Difference [95% CI]",adj=0,font=2,cex=.85) mtext("Tumor minus lung (log2 units)",side=1,line=2.2,at=1.6) ttl("C Cohort-specific effect estimates") },height=122) plate("Figure2",function(){ layout(matrix(c(1,2,3,3),2,byrow=TRUE),widths=c(1.15,1),heights=c(1.8,1)) z <- r[order(r$dependency,decreasing=TRUE),] y <- seq_len(nrow(z)) col <- ifelse(z$model_name=="NCI-H1836",accent, ifelse(z$model_name %in% c("NCI-H211","DMS 53"),pal[2],"grey54")) for(k in 1:2) { par(mar=c(3.8,if(k==1) 7.0 else 1.1,2.1,1.05)) v <- z[[c("chronos","dependency")[k]]] plot(v,y,type="n",ylim=c(.4,11.6),xlim=if(k==1) c(-.8,.06) else c(-.94,.16), yaxt="n",ylab="",xlab=c("Chronos gene effect","DEMETER2 score")[k]) if(k==1) axis(2,y,z$model_name,las=1,tick=FALSE,cex.axis=.83) abline(h=y,col="#ECEFF1",lwd=.55) abline(v=0,lty=2,col="grey55") segments(0,y,v,y,col=col) points(v,y,pch=ifelse(col=="grey54",16,18),col=col,cex=.93) ttl(c("A CRISPR knockout","B RNAi knockdown")[k]) } par(mar=c(3.5,5.0,2.1,1.05)) strip(list(g$chronos[g$group=="LUAD"],g$chronos[g$group=="SCLC"]),c("LUAD (47)","SCLC (23)"),"Chronos gene effect",-.5) ttl("C CRISPR effects across histologies") },height=139) plate("Figure3",function(){ u <- read("chan_umap.csv") stopifnot(nrow(u)==11744,all(u$detected %in% c(0,1))) layout(rbind(1:4,5:8,c(9,9,10,10)),heights=c(1,1,1.35)) ids <- unique(sort(u$patient)) cols <- c(malignant=pal[2],nonmalignant_epithelial=pal[1],immune="#5777A1", Fibroblast="#BB8B3F",Endothelial="#765896",other="#818181") ranges <- list(range(u$UMAP1),range(u$UMAP2)) for(kind in c("labels","detected")) for(i in seq_along(ids)){ z <- u[u$patient==ids[i],] par(mar=c(.9,if(i==1) 2.7 else .25,if(kind=="labels") 2 else .2,.25)) plot(NA,xlim=ranges[[1]],ylim=ranges[[2]],asp=1,axes=FALSE,xlab="",ylab="") box(col="#D8DDE1",lwd=.5) if(kind=="labels") points(z$UMAP1,z$UMAP2,pch=16,cex=.28,col=adjustcolor(cols[z$group],.62)) else { j <- z$detected==1 points(z$UMAP1[!j],z$UMAP2[!j],pch=16,cex=.24,col="#CAD0D5") points(z$UMAP1[j],z$UMAP2[j],pch=16,cex=.35,col="#A03951") } if(kind=="labels") title(paste("Patient",sub("HTA8_","",ids[i])),cex.main=.88) if(i==1) mtext(if(kind=="labels") "A Labels" else "B CIP2A",side=2,line=1.0,las=0,cex=.84) } par(mar=c(3.7,4.6,2.1,1.1)) z <- subset(p,stratum=="all") plot(NA,xlim=c(.72,2.28),ylim=c(0,52),xaxt="n",xlab="",ylab="CIP2A CPM") axis(1,1:2,c("Lung epithelium","Tumor group")) for(i in seq_len(nrow(z))) { k <- 1+(z$cohort[i]!="Chan") lines(1:2,c(z$cpm_nonmalignant_epithelial[i],z$cpm_malignant[i]),type="o",pch=c(16,17)[k],col=pal[k],lwd=1) } ttl("C Paired patient-level expression") legend("topleft",c("Chan","GSE281740"),col=pal,pch=c(16,17),lty=1,bty="n",cex=.8) par(mar=c(.8,1.1,2.0,.5)); plot.new(); plot.window(xlim=c(0,1),ylim=c(0,1)) ttl("D Cells contributing to C") text(c(.05,.57,.93),.94,c("Patient","Epithelium","Tumor"),adj=0,font=2,cex=.88) yy <- seq(.78,.43,length.out=nrow(z)) text(.05,yy,toupper(sub("HTA8_","",z$sample)),adj=0,cex=.88) text(.57,yy,z$cells_nonmalignant_epithelial,adj=0,cex=.88) text(.93,yy,z$cells_malignant,adj=1,cex=.88) legend(.02,.31,c("Tumor group","Lung epithelium","Immune","Fibroblast","Endothelial","Other"), col=unname(cols),pch=16,ncol=2,bty="n",cex=.73,y.intersp=1.1) text(.04,.03,"Panel B: burgundy = detected; gray = zero counts",adj=0,cex=.73) },height=159) plate("Figure4",function(){ layout(matrix(1:2,nrow=1),widths=c(1,1.13)) par(mar=c(4.25,4.35,2,1)) z <- subset(p,cohort=="GSE281740" & stratum %in% c("all","low_cycle")) plot(NA,xlim=c(.62,2.38),ylim=c(-4,48),xaxt="n",xlab="",ylab="Tumor minus epithelial (CPM)") axis(1,1:2,c("All cells","Low-cycle markers")) abline(h=0,lty=2,col="grey55") for(i in 1:2){ x <- z[z$sample==c("ru1334","ru1449a")[i],] lines(1:2,x$delta_cpm[match(c("all","low_cycle"),x$stratum)],type="o",col=pal[i],pch=c(16,17)[i],lwd=1.2) } legend("topright",c("RU1334","RU1449A"),col=pal,pch=c(16,17),lty=1,bty="n",cex=.85) ttl("A Effect after cycle stratification") q <- subset(read("gse281740_cnv_effects.csv"),stratum=="low_cycle_depth_matched") par(mar=c(2.4,1.05,2,1)); plot.new(); plot.window(xlim=c(0,1),ylim=c(0,1)) ttl("B Counts after low-cycle depth matching") heads <- c("Patient","Group","Cells","Total UMI","CIP2A UMI") xx <- c(.015,.23,.61,.80,.995) for(i in 1:5) text(xx[i],.93,heads[i],adj=if(i<=2) 0 else 1,font=2,cex=.81) segments(.01,.86,.99,.86,col="grey70") q <- q[order(q$sample,q$group),] for(i in seq_len(nrow(q))){ v <- q[i,] labels <- c(toupper(v$sample),if(v$group=="CNV_low_epithelial") "Lung epithelium" else "Tumor group", v$cells,format(v$umi,big.mark=","),v$cip2a) y <- .75-(i-1)*.16 for(j in 1:5) text(xx[j],y,labels[j],adj=if(j<=2) 0 else 1,cex=.84,font=if(j==5) 2 else 1) } },height=79) plate("GraphicalAbstract",function(){ layout(matrix(1:3,nrow=1),widths=c(1.2,1,1.05)) par(mar=c(2.9,.9,2.3,.85)) plot.new(); plot.window(xlim=c(0,1),ylim=c(0,1)) ttl("A Public data") y <- c(.79,.5,.21) rect(.05,y-.09,.95,y+.09,col=c("#E1EDEB","#F1E2E7","#E9E5F2"),border=NA) text(.5,y,c("Two bulk tissue cohorts","Two single-cell cohorts","CRISPR and RNAi screens"),cex=.88) par(mar=c(3.5,3.4,2.3,.8)) z <- subset(t,accession=="GSE149507") z <- reshape(z[c("patient","tissue","CIP2A")],idvar="patient",timevar="tissue",direction="wide") plot(NA,xlim=c(.7,2.3),ylim=range(z[-1]),xlab="",xaxt="n",ylab="CIP2A log2 expression") axis(1,1:2,c("Lung","Tumor")) segments(1,z$CIP2A.normal,2,z$CIP2A.tumor,col=adjustcolor(pal[1],.62)) ttl("B All 18 tissue pairs") par(mar=c(2.9,.7,2.3,.6)); plot.new(); plot.window(xlim=c(0,1),ylim=c(0,1)) ttl("C Two concordant models") text(c(.57,.89),.83,c("CRISPR","RNAi"),font=2,cex=.88) models <- c("NCI-H211","DMS 53") z <- r[match(models,r$model_name),] for(i in 1:2){ y <- c(.59,.34)[i] text(.01,y,models[i],adj=0,cex=.86) text(c(.57,.89),y,sprintf("%.2f",c(z$chronos[i],z$dependency[i])),cex=.95,col=pal[2],font=2) } },height=63)