Plotting spectroscopic data using ggplot2

Question

I am trying to plot spectroscopic data using ggplot2. I get my data in the following form:

data structure

My code so far is:

library(ggplot2)
library(reshape2)
melt_data <- melt(spectroscopic_data, id.vars = "sample_name", variable.name = "wavenumber", value.name = "intensity")
melt_data$probe = factor(melt_data$probe)
melt_data$wellenzahl = as.numeric(levels(melt_data$wellenzahl))[melt_data$wellenzahl]
ggplot(melt_data, aes(x=wavenumber, y=intensity, group=sample_name, color=sample_name)) + geom_line() +
scale_x_reverse(breaks=c(10000, 9500, 9000, 8500, 8000, 7500, 7000, 6500, 6000, 5500, 5000, 4500, 4000)) +
scale_color_manual(values=c("#FF0000", "#0000CC", "#00CC00", "#FF00FF", "#FF9900", "#000000", "#999900", "#33FFFF", "#FFCCFF", "#FFFF00", "#999999", "#9933FF", "#993300", "#99FF33")) + 
theme_bw() + 
theme(legend.position = "bottom") +
labs(x=expression(wavenumbers), y="intensity", colour = "") + 
theme(legend.text=element_text(size=10), axis.text=element_text(size=12), axis.title=element_text(size=14)) + 
guides(colour = guide_legend(ncol = 2, keywidth=1.5, keyheight=1, override.aes = list(size=1.8)))

I need the same color for aaa-samples, bbb-samples and so on (multiple measurements of one sample) but the plot does not work. I get a plot that looks like this when you zoom in:

zoom of current plot

It looks like ggplot2 connects two samples/lines of the same measurement instead of plotting them separately. Does anyone have an idea? I am trying to fix this since hours...

Thank you!

Welcome to Stack Overflow! This is a solid first question (I wish all the new questions were this quality). I'll just give you links, in case you want to take the tour. If you have more questions, you can head over to the help center as well. — Delioth
– Delioth, Commented Apr 17, 2017 at 21:27

an.be · Accepted Answer · 2017-04-30 13:30:44Z

Here is my result after Luke C's awesome support:

library(ggplot2)
library(reshape2)

melted_data <- melt(newtestdata, id.vars = c("sample_name", 
"sample_id"), variable.name = "wavenumber", value.name = "intensity")

melted_data$wavenumber=as.numeric(levels(melted_data$wavenumber))[melted_data$wavenumber]

ggplot(melted_data, aes(x=wavenumber, y=intensity, group = sample_id, color = sample_name)) + geom_line() +
scale_x_reverse(breaks=c(1005, 1200, 1400), expand = c(0.01, 0.01)) +  
scale_y_continuous(breaks=c(0, 0.5, 1.0, 1.5, 2.0), expand = c(0.01, 0.01)) +

scale_color_manual(values=c("#FF0000", "#0000CC", "#00CC00", "#FF00FF", "#FF9900", "#000000")) +

theme_bw() + 
theme(legend.position = "bottom") + 
theme(plot.margin=unit(c(1,1,0.5,1),"cm")) +

labs(x=expression(wavenumbers~"in"~cm^{"-1"}), y="absorbance in a.u.", colour = "") + 
theme(legend.text=element_text(size=10), axis.text=element_text(size=12), axis.title=element_text(size=14)) + 
guides(colour = guide_legend(ncol = 3, keywidth=1.5, keyheight=1, override.aes = list(size=1.2)))

ggsave("buechi-all.pdf", width = 11.69, height = 8.27)

Luke C · Accepted Answer · 2017-04-18 15:56:42Z

0

One way is to add a sample id to your data frame before you reshape it. That will allow you to keep the names like "aaa" and "bbb" but assign a unique identifier to act as your grouping variable (since it cannot differentiate between two observations at the same x variable otherwise). For an example where I tried to mimic your input data:

ex<-cbind(c("aaa","aaa","bbb","bbb"), c(0.426,0.405,0.409,0.395), c(0.430,0.408,0.411,0.399), c(0.432,0.411,0.413,0.401))

ex<- as.data.frame(ex)

colnames(ex) <- c("sample_name", "4000", "4004", "4008")

ex$sample_id<-1:nrow(ex)

melt <- melt(ex, id.vars = c("sample_name", "sample_id"), variable.name = "wavenumber", value.name = "intensity")

ggplot(melt, aes(x = wavenumber, y = intensity, group = sample_id, color = sample_name)) +
  geom_line() +
  theme_classic()

This outputs separate lines for different measurements of samples grouped by sample id, but keeping the color according to the sample name:

Is that sort of what you're after?

Edits below

To show the same approach with a larger dataset:

alpha <-rep(sapply(letters[1:10], function(x) {paste(x,x,x, sep = "")}), each = 2)
adf <- data.frame(alpha)
adf$sample_id <- seq(1, (length(alpha)))
adf$t <- rnorm(20, 0.4, 0.1)

wavenum <- seq(4, 1503)
for(i in wavenum){
  for(j in 1:length(alpha)){
    adf[j,i] <- adf[j,i-1] + (rnorm(1, 0.01, 0.01))
  }
}
adf[1:10, 1:10]

anames <- c("sample_name", "sample_id", (1400 + 4 * seq(0, 1500)))
names(adf)<-anames

melt <- melt(adf, id.vars = c("sample_name", "sample_id"), variable.name = "wavenumber", value.name = "intensity")

head(melt)

ggplot(melt[1:1500,], aes(x = wavenumber, y = intensity, group = sample_id, color = sample_name)) +
  geom_line(lwd = 1.5) +
  theme_classic()

This will give a similar plot to the one above, where each sample has an individual line for each measurement that are both the same color.

If I'm still missing what you're actually after, I apologize!

edited Apr 18, 2017 at 15:56

answered Apr 17, 2017 at 21:50

Luke C

10.4k1 gold badge17 silver badges24 bronze badges

4 Comments

an.be Over a year ago

Thank you for your reply! Since I am new to R and ggplot2, I am not sure - but I think your approach might be impractical because those datasets are huge (1500 wavenumbers for example). Maybe the problem is in the line melt_data$wavenumber = as.numeric(levels(melt_data$wavenumber))[melt_data$wavenumber] ?? Any idea?

Luke C Over a year ago

I don't think there is an actual problem with your code, unless you're getting an error- I think ggplot is just constrained by not knowing how to separate two samples with the same name. I'll edit my answer to show my approach working with a larger dataset, since as far as I know the size shouldn't matter in this case (beyond processing time).

an.be Over a year ago

Thanks again, I am going to try your approach with my data as soon as possible and I will let you know! Thank you so far!

an.be Over a year ago

Luke C, you are awesome!!! I tried it today - worked perfectly! Thank you very much!!! I will add an answer with my data set and result, it might help someone with less r-experience (like me)

Collectives™ on Stack Overflow

Plotting spectroscopic data using ggplot2

2 Answers 2

Comments

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related