The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
edited on 2016.01.22
Fluorescence fingerprint or more commonly known as complete fluorescence excitation-emission matrix (EEM) is a 3-dimensional data consisting of excitation, emission and intensity axis. The multi-dimension set EEM data apart from other signal processing. Thus, EEM package was developed to facilitate data analysis in R. Basic tools for importing raw data files, deleting Rayleigh scattering rays, unfolding 3-dimensional to 2-dimentional matrix (and vice versa) for further multivariate analysis, and visualizing data are provided in this package. The author has intended this package to be used as a bridge between raw data files and other analysis tools.
readEEM function can read raw data files into R. Currently the supported raw data files are limited to certain fluorescence spectrometers (see ?readEEM or impoting raw files vignette for more details). Basically readEEM will look for a certain keyword in raw files and start to read in the lines below them. Please send a word or pull request to add support for other formats.
Raw data files can be imported using any of the commands below. It can accept both file and folder path. Multiple paths can be specified using a vector.
library(EEM) # load library
# read raw data files from a file
data <- readEEM(file)
data <- readEEM("sample1.txt") # read in a file
data <-readEEM(c("sample1.txt", "sample2.txt")) # read in two files
# read raw data files from a folder
data <- readEEM(folder)
data <-readEEM("C:\\data") # full path. Note that the slash is doubled.
data <- readEEM("C:/data") # read in all files in data folder. Aside from double slashes,
# a reverted slash can also be used.
# read raw data files from the current working folder
setwd(choose.dir()) # set working folder interactively (only work in windows)
data <- readEEM(getwd()) # read raw data files in current working folder
The data is imported as a list and was given a class name of EEM. The original file names are can be retrieved using names(EEM), and additional information can be obtained using summary(data).
For demonstation purpose, a dataset called “applejuice” is attached with the package. It can be called by data(applejuice). More information about the dataset can be accessed by ?applejuice.
# load dataset
data(applejuice)
class(applejuice) # EEM class
## [1] "EEM"
# get sample names
names(applejuice)
## [1] "Aomori-Fuji-1-1" "Aomori-Fuji-1-2" "Aomori-Fuji-2-1"
## [4] "Aomori-Fuji-2-2" "Aomori-Jona-1-1" "Aomori-Jona-1-2"
## [7] "Aomori-Jona-2-1" "Aomori-Jona-2-2" "Aomori-Ohrin-1-1"
## [10] "Aomori-Ohrin-1-2" "Aomori-Ohrin-2-1" "Aomori-Ohrin-2-2"
## [13] "NZ-Envy-1-1" "NZ-Envy-1-2" "NZ-Envy-2-1"
## [16] "NZ-Envy-2-2" "NZ-Fuji-1-1" "NZ-Fuji-1-2"
## [19] "NZ-Fuji-2-1" "NZ-Fuji-2-2" "NZ-Jazz-1-1"
## [22] "NZ-Jazz-1-2" "NZ-Jazz-2-1" "NZ-Jazz-2-2"
# use summary to see information about the dataset.
summary(applejuice)
## Number of samples: 24
## Sample names:
## [1] "Aomori-Fuji-1-1" "Aomori-Fuji-1-2" "Aomori-Fuji-2-1"
## [4] "Aomori-Fuji-2-2" "Aomori-Jona-1-1" "Aomori-Jona-1-2"
## [7] "Aomori-Jona-2-1" "Aomori-Jona-2-2" "Aomori-Ohrin-1-1"
## [10] "Aomori-Ohrin-1-2" "Aomori-Ohrin-2-1" "Aomori-Ohrin-2-2"
## [13] "NZ-Envy-1-1" "NZ-Envy-1-2" "NZ-Envy-2-1"
## [16] "NZ-Envy-2-2" "NZ-Fuji-1-1" "NZ-Fuji-1-2"
## [19] "NZ-Fuji-2-1" "NZ-Fuji-2-2" "NZ-Jazz-1-1"
## [22] "NZ-Jazz-1-2" "NZ-Jazz-2-1" "NZ-Jazz-2-2"
## Dimension [EmxEx]: 48x26
## EX range: 200~450 [nm]
## EM range: 230~700 [nm]
EEM data is usually visualized using a contour representation. Two functions are available for creating contours.
drawEEM is built based on filled.contour of graphics package and can be used to draw any sample of an EEM class object.
# draw EEM of sample no.1
drawEEM(applejuice, n = 1)
# draw EEM of sample no.1 with different color
drawEEM(applejuice, n = 1, color.palette = cm.colors)
The axis can also be flipped by setting flipaxis to TRUE.
# flip the axis
drawEEM(applejuice, n = 1, flipaxis = TRUE)
drawEEMgg is an alternative contour creator. It uses ggplot2 package to draw contour.
# draw EEM of sample no.1
drawEEMgg(applejuice, n = 1)
# all functionalities in ggplot can be applied directly
library(ggplot2)
## Warning: package 'ggplot2' was built under R version 3.2.4
# add grid line to the plot
drawEEMgg(applejuice, n = 1) + theme(panel.grid = element_line(color = "grey"),
panel.grid.major = element_line(colour = "grey"))
Raw EEM data typically requires data cleaning, although some recent machines produced thoroughly cleaned data. Many papers (Fujita et al. (2010), Murphy et al. (2013)) have already discussed about the methods for cleaning and processing EEM data.
delScatteringThe Rayleign scattering rays of different orders can be deleted using delScattering. It is possible to choose whether to fill in the blank with NA or 0 by specifying rep (replace) argument. By running this function, the regions unrelated to fluorescence (where Em < Ex) will be also be deleted.
# delete scattering regions and assign them as NA
applejuice_delS <- delScattering(applejuice, rep = NA)
drawEEM(applejuice_delS, 1)
The width of each region to be deleted can also be set manually. The default values can be viewed through ?delScattering.
applejuice_delS <- delScattering(applejuice, rep = NA, first = 30, second = 0, third = 0, forth = 0)
drawEEM(applejuice_delS, 1)
rep was set to NA for demonstration purpose. However, since missing values cannot be included in some multivariate analysis, rep should be set to 0.
applejuice_delS <- delScattering(applejuice, rep = 0,
first = 30, second = 0, third = 0, forth = 0)
delScattering2delScattering2 is different from delScattering in the way that the region beyond
the second-order scattering ray is removed, in addition to the regions where emission wavelength is shorten than excitation light (Em <= Ex) and the first- and second-order scattering rays.
drawEEM(delScattering2(applejuice, NA), 1)
cutEEM function offers a method to cut portions of EEM by specifying cutEX and cutEM argument values. However, please take note that it is not possible to cut portion in the middle.
applejuice_delS_cut <- cutEEM(applejuice_delS, cutEX = 350:500, cutEM = 500:700)
drawEEM(applejuice_delS_cut, 1)
EEM data can be unfolded into a matrix with columns as variables (wavelength conditions) and rows as samples, which is a common format for multivariate analysis.
## unfold EEM into EEM_uf (matrix form with samples x variables dimension)
applejuice_delS_uf <- unfold(applejuice_delS)
# dimension of unfolded data
dim(applejuice_delS_uf)
## [1] 24 1248
# take a look at unfolded data
applejuice_delS_uf[1:5 ,1:5]
## EX200EM230 EX200EM240 EX200EM250 EX200EM260 EX200EM270
## Aomori-Fuji-1-1 1.584740 2.750280 2.71587 6.51603 21.1658
## Aomori-Fuji-1-2 0.480644 2.064270 4.72847 5.65419 27.9555
## Aomori-Fuji-2-1 2.430540 1.427630 3.28927 5.36893 23.0553
## Aomori-Fuji-2-2 0.461552 2.157030 3.26636 5.78505 28.7779
## Aomori-Jona-1-1 -0.170955 0.362999 2.46317 6.07851 22.4464
Unfolded data can also be folded back into EEM class by fold function.
Unfolded data can be normalized using normalize function to adjust the scaling difference, which is a common bias in spectroscopic applications. This difference can be caused by the scattering effect, source/detector variation and instrumental sensitivity. Normalize function will do the row processing of the unfolded data by divide each variable by the sum of the absolute value of all variables for the given sample. The output will return a matrix where each row is a vector with unit area (area = 1).
# normalize data
applejuice_delS_uf_norm <- normalize(applejuice_delS_uf)
# the absolute sum of each row should equal to 1
rowSums(abs(applejuice_delS_uf_norm))
## Aomori-Fuji-1-1 Aomori-Fuji-1-2 Aomori-Fuji-2-1 Aomori-Fuji-2-2
## 1 1 1 1
## Aomori-Jona-1-1 Aomori-Jona-1-2 Aomori-Jona-2-1 Aomori-Jona-2-2
## 1 1 1 1
## Aomori-Ohrin-1-1 Aomori-Ohrin-1-2 Aomori-Ohrin-2-1 Aomori-Ohrin-2-2
## 1 1 1 1
## NZ-Envy-1-1 NZ-Envy-1-2 NZ-Envy-2-1 NZ-Envy-2-2
## 1 1 1 1
## NZ-Fuji-1-1 NZ-Fuji-1-2 NZ-Fuji-2-1 NZ-Fuji-2-2
## 1 1 1 1
## NZ-Jazz-1-1 NZ-Jazz-1-2 NZ-Jazz-2-1 NZ-Jazz-2-2
## 1 1 1 1
After preprocessing and unfolding, the data matrix can be exported as text files to be analyzed by other stats softwares. Note that further analysis is also possible in R.
# export as csv file
write.csv(applejuice_delS_uf, "applejuice.csv")
After analysis with other softwares, the returned values such as regression coefficient or VIP value can be imported back in R so that contours can be easily plotted. The detailed walk through can be found in impoting raw files vignette for more details)
prcomp of stats package can be used to perform PCA on the unfolded data.
# perform PCA
result <- prcomp(applejuice_delS_uf_norm) # mean-centering is enabled by default
# plot scree plot
screeplot(result, npcs = 10, type = "lines", main = "Screeplot")
The score and loading can be plotted by plotScore and plotLoading, respectively.
# plot score plot
plotScore(result, xPC = 1, yPC = 2) # pc 1 vs pc 2
# plot loading plot
plotLoading(result, ncomp = 1) # loading 1
For our example, PCA will be used to test whether EEM can discriminate apples of different production area and cultivars. Since those information is hidden in the sample names, they will be retrieved first.
# extract sample name
sName <- names(applejuice)
# country of apple production
country <- sapply(strsplit(sName, split = "-"), "[", 1)
table(country) # counts of samples grouped by country
## country
## Aomori NZ
## 12 12
# cultivar of apples
cultivar <- sapply(strsplit(sName, split = "-"), "[", 2)
table(cultivar) # counts of samples grouped by cultivar
## cultivar
## Envy Fuji Jazz Jona Ohrin
## 4 8 4 4 4
To plot score plot with points colored by group, plotScore and plotScorem can be used.
# plot score plot with grouping
plotScore(result, xPC = 1, yPC = 2,country, legendlocation = "topright")
# plot score using scatterplot matrix with grouping
plotScorem(result, ncomp = 5, country)
plotScorem(result, ncomp = 5, cultivar, cex = 1)
PLS regression can be calculated using plsr function of pls package. plsr function returns an output variable of the class mvr. The latent variables can be visualized in a contour representation using plotLoading function. Similarly, the regression coefficient can be visualized in a contour representation using plotReg function.
# load gluten data
data(gluten)
gluten_uf <- unfold(gluten) # unfold list into matrix
# delete columns with NA values
index <- colSums(is.na(gluten_uf)) == 0
gluten_uf <- gluten_uf[, index]
gluten_ratio <- as.numeric(names(gluten))
require(pls)
model <- plsr(gluten_ratio ~ gluten_uf, ncomp = 3)
plotLoading(model, ncomp = 3)
plotReg(model)
Fujita, K., Tsuta, M., Kokawa, M., & Sugiyama, J. (2010). Detection of deoxynivalenol using fluorescence excitation-emission matrix. Food and Bioprocess Technology, 3(6), 922-927.
Murphy, K. R., Stedmon, C. A., Graeber, D., & Bro, R. (2013). Tutorial Review: Fluorescence spectroscopy and multi-way techniques. PARAFAC. Analytical Methods.
These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.