pca_by_var: pca_by_var

In OxfordCMS/OCMSutility: Utility functions for OCMS

pca_by_var

Description

Overlay numeric metadata variables onto a PCA score plot. Useful during exploratory analysis

Usage

pca_by_var(
  ddata,
  mdata,
  PC = c(1, 2),
  biplot = TRUE,
  score_colour = FALSE,
  score_label = FALSE
)

Arguments

ddata	dataframe. samples in rows. sample ids in rownames. rows should match mdata
mdata	dataframe. samples in rows. sample ids in rownames. rows should match ddata. Variables can be numeric, characters, or factors
PC	numeric vector of length 2. specifies the PCs to plot. default is c(1,2)
biplot	logical. show PCA as score plot or biplot
score_colour	string. colour data points on score/biplot. should be column name in mdata. When set to FALSE, data points are black. Default FALSE
score_label	string. label data ponts on score/biplot. should be column name in mdata. When set to FALSE, data ponts are not labelled. Default FALSE

Value

named list of plots. First plot in list is main_pca. Subsequent plots in list correspond to metadata variables (columns in metadata)

Examples

set.seed(1)
data(dss_example)
ddata <- dss_example$merged_abundance_id[,2:26]
rownames(ddata) <- dss_example$merged_abundance_id[,1]
ddata <- as.data.frame(t(ddata))

mdata <- dss_example$metadata
mdata <- mdata[match(rownames(ddata), mdata$sampleID),]

# creating some dummy metadata variable
mdata$var1 <- rep(rnorm(5, 25, 3), each=5)
mdata$var2 <- rep(rnorm(5, 3, 0.5), 5)
mdata$var3 <- as.factor(rep(letters[1:5], each=5))
mdata <- mdata[,c('Phenotype','var1','var2','var3')]
p_list <- pca_by_var(ddata, mdata)

# biplot
p_list$main_pca
# pca with metadata variables overlayed
p_list$Phenotype
p_list$var1
p_list$var2
p_list$var3

# can use cowplot::plot_grid to put all plots into one
cowplot::plot_grid(plotlist=list(p_list$Phenotype, p_list$var1, p_list$var2, p_list$var3))

OxfordCMS/OCMSutility documentation built on Feb. 27, 2025, 8:19 p.m.