First we need to load the package.
# load glycanr package
library(glycanr)
## From version 0.3 functions tanorm and glyco.outliers expect data frames
## in long format.
Lets create a data.frame to simulate our glycan data.
set.seed(123)
n <- 200
X <- data.frame(ID=1:n, GP1=runif(n), GP2=rexp(n, 0.3),
GP3=rgamma(n, 2), cc=factor(sample(1:2, n, replace=TRUE)))
Now we have data.frame X where GP represents glycans, ID represents e.g. sample IDs and cc represents Case/Control status.
head(X)
## ID GP1 GP2 GP3 cc
## 1 1 0.2875775 6.0104311 2.4551743 1
## 2 2 0.7883051 0.1001982 0.2778425 1
## 3 3 0.4089769 4.3448018 1.1637760 2
## 4 4 0.8830174 0.6659741 1.7217277 2
## 5 5 0.9404673 5.8381875 1.5926956 1
## 6 6 0.0455565 5.8788944 2.0702267 1
This data can now be plotted with glyco.plot function.
Basic usage is given as follows.
glyco.plot(X)
It plots boxplots for every column whose name starts with 'GP'. To change boxplots with violin plots (represents the density of a data) option violin should be used.
glyco.plot(X, violin=TRUE)
To separate boxplots (or violin plots) into different layers option collapse should be used.
glyco.plot(X, collapse=FALSE)
To plot log transformed data option log.transform should be used
glyco.plot(X, collapse=FALSE, log.transform=TRUE)
If you want to see the difference between groups with boxplots (or violin plots) option group should be used. It takes a character string representing the name of the column on grouping should be done.
Grouping by cc variable can be done like this.
glyco.plot(X, collapse=FALSE, log.transform=TRUE, group="cc")
## $p.val.unadj
## GP1 GP2 GP3
## 0.07474019 0.20680798 0.95699301
##
## $p.val
## GP1 GP2 GP3
## 0.2242206 0.4136160 0.9569930
##
## $plot
As it can be seen the default option is to also conduct a test in difference between groups (Mann-Whitney-Wilcoxon for 2 groups, Kruskal-Wallis for more groups) and print the obtained p-values. The printed values are corrected for multiple testing before printing. As the output you get original p-values and adjusted p-values together with the plot.
Method for multiple testing correction can be adjusted by parameter p.adjust.method.
glyco.plot(X, collapse=FALSE, log.transform=TRUE, group="cc", p.adjust.method="fdr")
## $p.val.unadj
## GP1 GP2 GP3
## 0.07474019 0.20680798 0.95699301
##
## $p.val
## GP1 GP2 GP3
## 0.2242206 0.3102120 0.9569930
##
## $plot
Printing p-values in plots can be omitted with print.p.values parameter.
glyco.plot(X, collapse=FALSE, log.transform=TRUE, group="cc", p.adjust.method="fdr",
print.p.values=FALSE)
## $p.val.unadj
## GP1 GP2 GP3
## 0.07474019 0.20680798 0.95699301
##
## $p.val
## GP1 GP2 GP3
## 0.2242206 0.3102120 0.9569930
##
## $plot
When grouping, by default, all glycans are plotted. To plot only those that differ statistically significant parameter all should be used.
Function glyco.plot plots all columns whose name starts with GP. Since these plotting techniques can be used on other data as well there is the parameter glyco.names to choose which columns you want to use.
glyco.plot(X, collapse=FALSE, log.transform=TRUE, group="cc", p.adjust.method="fdr",
print.p.values=FALSE, glyco.names=c("GP1", "GP2"))
## $p.val.unadj
## GP1 GP2
## 0.07474019 0.20680798
##
## $p.val
## GP1 GP2
## 0.1494804 0.2068080
##
## $plot
