Add a line to an existing bar graph and a second y axis

Question

I am new to R and stack overflow, so please bear with me. None of the other questions on here efficiently address what I'm after.

I have compared the abilities of four targeted gene panels to detect mutations in one standard (AcroMetrix). The mutations are of different types and frequencies, and I can easily generate a barplot accordingly. However, the panels don't target all mutations in AcroMetrix. So I would like to add a line of the percent targeted bases, with a concomitant y-axis on the right hand side.

Please see below for clarity.

# This generates the barplot

df<-data.frame(row.names=c("AcroMetrix","PV1_PV2","CHIPv2","TSACP","TSTP"),Germline=c(34,33,14,22,12),Somatic_5_15=c(341,331,281,249,147),Somatic_15_30=c(180,176,129,124,108))

df$name<-row.names(df)  
df_molten<-melt(df)
df_molten$name <-factor(df_molten$name,
levels = c("AcroMetrix","PV1_PV2","CHIPv2","TSACP","TSTP"))

    ggplot(df_molten,aes(x=name,y=value,fill=variable))+
geom_bar(stat='identity')+
scale_fill_discrete(labels=c("Germline","Somatic 5-15% VAF","Somatic 15-30% VAF"))+
geom_text(aes(label=value),size=3,fontface='bold',position=position_stack(vjust=.5))+
xlab("Panel")+
ylab("Counts")+
theme_bw()+
theme(panel.grid.major=element_blank(),
      panel.grid.minor=element_blank(),
      panel.background=element_blank(),
      axis.line=element_line(colour="black"),
      panel.border=element_blank(),
      legend.title=element_blank())

# The second set of data for the percent targets are as follows, and this needs to form the line graph and be compared to the Y axis on the right:
 df1 <- data.frame(row.names=c("AcroMetrix","PV1_PV2","CHIPv2","TSACP","TSTP"),Percent_targeted=c(100,100,77,73,49))

What count would correspont to 100%? An secondary y-axis (see [here](https://stackoverflow.com/questions/55648536/adding-a-line-independent-of-y-to-the-axis-in-r/55649884?noredirect=1#comment97989826_55649884)) only seems to make sense if 100% corresponds to the same count for each panel, right? — Simon, Apr 18 '19 at 13:57

Simon · Accepted Answer · 2019-04-18T23:24:11.823

Edit 3:

Sorry, just saw your sketch right now...

Use geom_point()and geom_line() to create the line and the points. Adding a number to Percent_targeted_scaled (1 in this example) moves points and lines up with respect to the bars. Change this until you have your inteded position.

Use size in geom_point() and lwd in geom_line() to create the appropriate point size and linewidth.

library(ggplot2)
library(reshape2)
library(scales)
df<-data.frame(row.names=c("AcroMetrix","PV1_PV2","CHIPv2","TSACP","TSTP"),Germline=c(34,33,14,22,12),Somatic_5_15=c(341,331,281,249,147),Somatic_15_30=c(180,176,129,124,108))

df$name<-row.names(df)

df_molten<-melt(df)

df_molten$name<-factor(df_molten$name,levels=c('AcroMetrix','PV1_PV2','CHIPv2','TSACP','TSTP'))
df_molten$Percent_targeted <- unlist(lapply(1:length(levels(df_molten$variable)), function(i){c(100,100,77,73,49)}))
# counts <- df_molten %>% group_by(name) %>% summarise(sum=round(sum(value)))
# df_molten$Percent_targeted <- round(unlist(lapply(1:length(levels(df_molten$variable)), function(i){counts$sum/counts$sum[1]})), 2)*100

gg <- ggplot(df_molten,aes(x=name,y=value,fill=variable))+
  geom_bar(stat='identity', width=.6)+
  scale_fill_discrete(labels=c("Germline","Somatic 5-15% VAF","Somatic 15-30% VAF"))+
  geom_text(aes(label=value),size=3,fontface='bold',position=position_stack(vjust=.5))+
  xlab("Panel")+ylab("Counts")+
  theme_bw()+
  theme(panel.grid.major=element_blank(),panel.grid.minor=element_blank(),panel.background=element_blank(),axis.line=element_line(colour="black"),panel.border=element_blank(),legend.title=element_blank())
gg <- gg + scale_y_continuous(expand = expand_scale(mult=c(0, 0.0)))


# get the sacle values of the current y-axis
gb <- ggplot_build(gg)
y.range <- gb$layout$panel_params[[1]]$y.range
y2.range <- range(df_molten$Percent_targeted)# extendrange(, f=0.01)
scale_factor <- (diff(y.range)/max(y2.range))
trans <- ~ ((. -y.range[1])/scale_factor)

df_molten$Percent_targeted_scaled <- rescale(df_molten$Percent_targeted, y.range, c(0, y2.range[2]))
df_molten$x <- which(levels(df_molten$name)%in%df_molten$name)#-.3

# gg <- gg + geom_segment(aes(x=x, xend=x, yend=Percent_targeted_scaled), y=0, size=2, data=df_molten)
# gg <- gg + geom_label(aes(label=paste0(Percent_targeted, '%'), x=x, y=Percent_targeted_scaled), fill='white', data=df_molten)
# gg <- gg + geom_hline(yintercept = y.range[2], linetype='longdash')
# gg <- gg + geom_label(aes(label=paste0(Percent_targeted, '%'), x=x, y=Percent_targeted_scaled), fill='white', data=df_molten, vjust=0)
gg <- gg + geom_point(aes(x=x, y=Percent_targeted_scaled+2), data=df_molten, show.legend = F, size=3)
gg <- gg + geom_line(aes(x=x, y=Percent_targeted_scaled+2), data=df_molten, lwd=1.5)
gg <- gg + scale_y_continuous(expand=expand_scale(mult=c(.05, .05)), sec.axis = sec_axis(trans, name = paste0("Percent genes targeted on ", levels(df_molten$name)[1]), labels = scales::percent(seq(0, 1, length.out = 5), scale=100)))
gg

Edit 2:

To get the points (or labels) to their percentage use the rescaled percentage value as y-aesthatic:

library(ggplot2)
library(reshape2)
library(scales)
df<-data.frame(row.names=c("AcroMetrix","PV1_PV2","CHIPv2","TSACP","TSTP"),Germline=c(34,33,14,22,12),Somatic_5_15=c(341,331,281,249,147),Somatic_15_30=c(180,176,129,124,108))

df$name<-row.names(df)

df_molten<-melt(df)

df_molten$name<-factor(df_molten$name,levels=c('AcroMetrix','PV1_PV2','CHIPv2','TSACP','TSTP'))
df_molten$Percent_targeted <- unlist(lapply(1:length(levels(df_molten$variable)), function(i){c(100,100,77,73,49)}))

gg <- ggplot(df_molten,aes(x=name,y=value,fill=variable))+
  geom_bar(stat='identity', width=.6)+
  scale_fill_discrete(labels=c("Germline","Somatic 5-15% VAF","Somatic 15-30% VAF"))+
  geom_text(aes(label=value),size=3,fontface='bold',position=position_stack(vjust=.5))+
  xlab("Panel")+ylab("Counts")+
  theme_bw()+
  theme(panel.grid.major=element_blank(),panel.grid.minor=element_blank(),panel.background=element_blank(),axis.line=element_line(colour="black"),panel.border=element_blank(),legend.title=element_blank())
gg <- gg + scale_y_continuous(expand = expand_scale(mult=c(0, 0.0)))


# get the sacle values of the current y-axis
gb <- ggplot_build(gg)
y.range <- gb$layout$panel_params[[1]]$y.range
y2.range <- range(df_molten$Percent_targeted)# extendrange(, f=0.01)
scale_factor <- (diff(y.range)/max(y2.range))
trans <- ~ ((. -y.range[1])/scale_factor)

df_molten$Percent_targeted_scaled <- rescale(df_molten$Percent_targeted, y.range, c(0, y2.range[2]))
df_molten$x <- which(levels(df_molten$name)%in%df_molten$name)#-.3

# gg <- gg + geom_segment(aes(x=x, xend=x, yend=Percent_targeted_scaled), y=0, size=2, data=df_molten)
# gg <- gg + geom_label(aes(label=paste0(Percent_targeted, '%'), x=x, y=Percent_targeted_scaled), fill='white', data=df_molten)
gg <- gg + geom_hline(yintercept = y.range[2], linetype='longdash')
gg <- gg + geom_label(aes(label=paste0(Percent_targeted, '%'), x=x, y=Percent_targeted_scaled), fill='white', data=df_molten, vjust=0)

gg <- gg + scale_y_continuous(expand=expand_scale(mult=c(.05, .05)), sec.axis = sec_axis(trans, name = paste0("Percent genes targeted on ", levels(df_molten$name)[1]), labels = scales::percent(seq(0, 1, length.out = 5), scale=100)))
gg

Edit:

I understand that the aim is to have a horizontal line at 100% which corresponds to the max on AcroMetrix.

So do you mean something like this:

library(ggplot2)
library(reshape2)
library(scales)
df<-data.frame(row.names=c("AcroMetrix","PV1_PV2","CHIPv2","TSACP","TSTP"),Germline=c(34,33,14,22,12),Somatic_5_15=c(341,331,281,249,147),Somatic_15_30=c(180,176,129,124,108))

df$name<-row.names(df)

df_molten<-melt(df)

df_molten$name<-factor(df_molten$name,levels=c('AcroMetrix','PV1_PV2','CHIPv2','TSACP','TSTP'))
df_molten$Percent_targeted <- unlist(lapply(1:length(levels(df_molten$variable)), function(i){c(100,100,77,73,49)}))

gg <- ggplot(df_molten,aes(x=name,y=value,fill=variable))+
  geom_bar(stat='identity', width=.6)+
  scale_fill_discrete(labels=c("Germline","Somatic 5-15% VAF","Somatic 15-30% VAF"))+
  geom_text(aes(label=value),size=3,fontface='bold',position=position_stack(vjust=.5))+
  xlab("Panel")+ylab("Counts")+
  theme_bw()+
  theme(panel.grid.major=element_blank(),panel.grid.minor=element_blank(),panel.background=element_blank(),axis.line=element_line(colour="black"),panel.border=element_blank(),legend.title=element_blank())
gg <- gg + scale_y_continuous(expand = expand_scale(mult=c(0, 0.0)))


# get the sacle values of the current y-axis
gb <- ggplot_build(gg)
y.range <- gb$layout$panel_params[[1]]$y.range
y2.range <- range(df_molten$Percent_targeted)# extendrange(, f=0.01)
scale_factor <- (diff(y.range)/max(y2.range))
trans <- ~ ((. -y.range[1])/scale_factor)

df_molten$Percent_targeted_scaled <- rescale(df_molten$Percent_targeted, y.range, c(0, y2.range[2]))
df_molten$x <- which(levels(df_molten$name)%in%df_molten$name)#-.3

# gg <- gg + geom_segment(aes(x=x, xend=x, yend=Percent_targeted_scaled), y=0, size=2, data=df_molten)
# gg <- gg + geom_label(aes(label=paste0(Percent_targeted, '%'), x=x, y=Percent_targeted_scaled), fill='white', data=df_molten)
gg <- gg + geom_hline(yintercept = y.range[2], linetype='longdash')
gg <- gg + geom_label(aes(label=paste0(Percent_targeted, '%'), x=x, y=y.range[2]+5), fill='white', data=df_molten, vjust=0)

gg <- gg + scale_y_continuous(expand=expand_scale(mult=c(.05, .05)), sec.axis = sec_axis(trans, name = paste0("Percent genes targeted on ", levels(df_molten$name)[1]), labels = scales::percent(seq(0, 1, length.out = 5), scale=100)))
gg

Original answer:

From the data you provide it seems to me that 100% is not the same on each panel.

However, you can do the requested like this:

library(ggplot2)
library(reshape2)
library(scales)
df<-data.frame(row.names=c("AcroMetrix","PV1_PV2","CHIPv2","TSACP","TSTP"),Germline=c(34,33,14,22,12),Somatic_5_15=c(341,331,281,249,147),Somatic_15_30=c(180,176,129,124,108))

df$name<-row.names(df)

df_molten<-melt(df)

df_molten$name<-factor(df_molten$name,levels=c('AcroMetrix','PV1_PV2','CHIPv2','TSACP','TSTP'))
df_molten$Percent_targeted <- unlist(lapply(1:length(levels(df_molten$variable)), function(i){c(100,100,77,73,49)}))

gg <- ggplot(df_molten,aes(x=name,y=value,fill=variable))+
  geom_bar(stat='identity', width=.6)+
  scale_fill_discrete(labels=c("Germline","Somatic 5-15% VAF","Somatic 15-30% VAF"))+
  geom_text(aes(label=value),size=3,fontface='bold',position=position_stack(vjust=.5))+
  xlab("Panel")+ylab("Counts")+
  theme_bw()+
  theme(panel.grid.major=element_blank(),panel.grid.minor=element_blank(),panel.background=element_blank(),axis.line=element_line(colour="black"),panel.border=element_blank(),legend.title=element_blank())
gg <- gg + scale_y_continuous(expand = expand_scale(mult=c(0, 0.0)))


# get the sacle values of the current y-axis
gb <- ggplot_build(gg)
y.range <- gb$layout$panel_params[[1]]$y.range
y2.range <- range(df_molten$Percent_targeted)# extendrange(, f=0.01)
scale_factor <- (diff(y.range)/max(y2.range))
trans <- ~ ((. -y.range[1])/scale_factor)

df_molten$Percent_targeted_scaled <- rescale(df_molten$Percent_targeted, y.range, c(0, y2.range[2]))
df_molten$x <- which(levels(df_molten$name)%in%df_molten$name)-.3

gg <- gg + geom_segment(aes(x=x, xend=x, yend=Percent_targeted_scaled), y=0, size=2, data=df_molten)
gg <- gg + geom_label(aes(label=paste0(Percent_targeted, '%'), x=x, y=Percent_targeted_scaled), fill='white', data=df_molten)
gg <- gg + scale_y_continuous(expand=expand_scale(mult=c(.05, .05)), sec.axis = sec_axis(trans, name = "Percent targeted", labels = scales::percent(seq(0, 1, length.out = 5), scale=100)))
gg

Hi Simon, thanks for this. This is close to what I want, but I wonder if we can have a single line run along the top of the graph with a point above each bar plot. I'll try construct an example plot shortly. AcroMetrix is the panel I'm comparing the others to. So PV1_PV2 for example, targets 100% of the genes on AcroMetrix, but I don't detect them all. That's what I'm showing for that panel and the other three, basically there's a gap between what's targeted and what's detected. I'd just like that line to run along the top of the bars, with a point above the centre of each bar if possible — MJ_Macca, Apr 18 '19 at 21:09
Hi Simon, many thanks again! Yes that’s closer. But those last three points (77, 73 and 49) should drop to their respective levels according to the right hand y axis. For example, they’re all sitting at 100% at the moment. I’ll put up an example in about 10-15 mins. Thanks so much for your help — MJ_Macca, Apr 18 '19 at 22:32
Hi Simon, that new edit looks good. I've put a sketch at the top of the page of what I'm thinking if that's doable. It would be good to keep the percent values above the points indicated on the line above each bar. I forgot to add that to the sketch! — MJ_Macca, Apr 18 '19 at 22:54
Hi Simon, that's fantastic, thanks so much, it is greatly appreciated! To bring the value labels back in above the points, presumably I just comment back in the gg+geom_label command. Thanks very much! — MJ_Macca, Apr 18 '19 at 23:21
Indeed, uncomment the `gg+ geom_label(...)` to get the labels back, but don't forget to add the same value as in `geom_line()` to `Percent_targeted_scaled` to have them on the respecive height. Feel free to mark the answer as correct if this fullfills your needs! Thank you! — Simon, Apr 18 '19 at 23:30
Thanks again Simon, that’s been most helpful and is greatly appreciated. I’ve marked the question as answered :) — MJ_Macca, Apr 19 '19 at 00:31

Add a line to an existing bar graph and a second y axis

1 Answers1