So this is a bit of a hack but it might work for you.
I introduce a third column in the graph to hold the labels as per my original post.
I pre-process your data to try and spread out the labels in this third column around the Tissue variable to that they don't appear all on top of each other.
my pre-processing is pretty ugly but works ok. Note that I only catered for a max of 4 cell.types as per your comment.
It gives me this graph:
My code:
data = data.frame("Tissue"=c("Adrenal gland", "Appendix", "Appendix"), "protein.expression" = c("No detect","No detect", "Medium"), "cell.type" = c("Glandular cells" ,"Lymphoid tissu","Glandular cells"))
# Pre-processing section.
# Step 1: find out the n of cell.types per tissue type
counters <- data %>% group_by(Tissue) %>% summarise(count = n())
# Step 2: Join n back to original data. Transform protein.expression to ordered factor
data <- data %>%
inner_join(counters, by="Tissue") %>%
mutate(protein = factor(protein.expression, levels=unique(protein.expression, decreasing = F), ordered=TRUE),
positionTissue = as.numeric(Tissue))
results <- data.frame()
# Step 3: Spread the cell.type labels around the position of the Tissue. 4 scenarios catered for.
for(t in unique(data$Tissue)){
subData <- filter(data, Tissue == t)
subData$spreader <- as.numeric(subData$Tissue)
if(length(unique(subData$cell.type)) == 2){
subData <- subData %>%
mutate(x=factor(cell.type, levels=unique(cell.type, decreasing = F),ordered=TRUE),
spreader = ifelse(as.numeric(x)==1,as.numeric(Tissue)-0.1,as.numeric(Tissue)+0.1)) %>%
select(-x)
results <- rbind(results, subData)
} else if(length(unique(subData$cell.type)) == 3){
subData <- subData %>%
mutate(x=factor(cell.type, levels=unique(cell.type, decreasing = F),ordered=TRUE),
spreader = ifelse(as.numeric(x)==1,as.numeric(Tissue)-0.15,
ifelse(as.numeric(x)==3,as.numeric(Tissue)+0.15,as.numeric(Tissue)))) %>%
select(-x)
results <- rbind(results, subData)
} else if(length(unique(subData$cell.type)) == 4){
subData <- subData %>%
mutate(x=factor(cell.type, levels=unique(cell.type, decreasing = F),ordered=TRUE),
spreader = ifelse(as.numeric(x)==1,as.numeric(Tissue)-0.2,
ifelse(as.numeric(x)==2,as.numeric(Tissue)-0.1,
ifelse(as.numeric(x)==3,as.numeric(Tissue)+0.1,
ifelse(as.numeric(x)==4,as.numeric(Tissue)+0.2,as.numeric(Tissue)))))) %>%
select(-x)
results <- rbind(results, subData)
} else{
results <- rbind(results, subData)
}
}
# Plot the data based on the new label position "spreader" variable
ggplot(results, aes(x = positionTissue, y = protein, label=cell.type)) +
geom_point(stat='identity', aes(col=protein.expression), size=12) +
geom_text(aes(y=0.5,label=Tissue), size=8, fontface="bold", angle=90)+
geom_label(aes(y="zzz", x=spreader, fill=protein), colour="white") +
theme_classic() +
scale_x_continuous(limits = c(min(as.numeric(data$Tissue))-0.5,max(as.numeric(data$Tissue))+0.5))+
scale_y_discrete(breaks=c("Medium","No detect")) +
labs(title="Protein Atlas") +
guides(fill=guide_legend(title="Protein expression"))+
ylab("Cell types measured per tissue") +
xlab("") +
#ylim(1,4) +
coord_flip()+
theme(axis.text.x = element_text(size = 25),
axis.text.y = element_text(colour = NA),
legend.position = "none",
axis.title.x = element_text(size=30),
axis.title.y = element_text(size = 30, margin = margin(t = 0, r = 20, b = 0, l = 0)),
legend.title = element_text(size = 30),
legend.text = element_text(size = 25),
legend.key.size = unit(2, 'cm'),
axis.ticks.length=unit(.01, "cm"),
strip.text.y = element_text(angle = 0))
Edit #2:
Update to retain label colours by creating n positions where n is the number of cell.types:
data = data %>%
mutate(position = paste("z",cell.type))
Then you can use this new position variable instead of the static "zzz" I suggested in my original post. Your labels will have the correct colours, but your chart will look odd if there are a lot of cell.types.
geom_label(aes(y=position, label = cell.type)) +
EDIT #1: Update to avoid overlapping labels by grouping cell.types to a single label per tissue.
Creating a new label field that concatenates the individual labels for each tissue type:
data = data %>%
group_by(Tissue) %>%
mutate(label = paste(cell.type, collapse = "; "))
And amend the ggplot call to use this new field instead of the existing cell.type field:
geom_text(aes(y="zzz", label = label), size = 6, fontface = "bold", colour = "white")+
or:
geom_label(aes(y="zzz", label = label),) +
ORIGINAL POST:
You could plot your labels at a third position (e.g. "zzz") and then hide that position from the set of axis labels using scale_x_discrete(breaks=c()).
ggplot(data, aes(x = Tissue, y = factor(protein.expression,
levels=unique(protein.expression,
decreasing = F),
ordered=TRUE), fill = protein.expression,
label = cell.type))+
geom_point(stat='identity', aes(col=protein.expression), size=12) +
geom_text(aes(y="zzz"), size = 6, fontface = "bold", colour = "white")+
geom_label(aes(y="zzz"),) +
# facet_grid(cell.type ~ ., scales = "free", space = "free") +
# scale_fill_manual(values = myPalette, drop = FALSE) +
# scale_color_manual(values = myPalette, drop = FALSE) +
theme_classic() +
scale_y_discrete(breaks=c("Medium","No detect"))+
labs(title="Protein Atlas") +
guides(fill=guide_legend(title="Protein expression"))+
ylab("Cell types measured per tissue") +
#ylim(1,4) +
coord_flip()+
theme(axis.text.x = element_text(size = 25, vjust = 0.5, hjust = .9),
axis.text.y = element_text(size = 25),
legend.position = "none",
axis.title.x = element_text(size=30),
axis.title.y = element_text(size = 30, margin = margin(t = 0, r = 20, b = 0, l = 0)),
legend.title = element_text(size = 30),
legend.text = element_text(size = 25),
legend.key.size = unit(2, 'cm'),
axis.ticks.length=unit(.01, "cm"),
strip.text.y = element_text(angle = 0))
