Analysing phenotype microarray data: drawing principal components

Using BiodiversityR and opm to generate PCA bi-plots of estimated parameters. An advantage of such plots is that the overall differences between the organisms can be plotted together with indications of the effect of certain substrates on these differences. Furthermore, the data are reduced to their independent components before plotting. Depending on the biological question, this might be more interesting than drawing heat maps. Some BiodiversityR functions do not always seem to work as intended, see the comments below.

Author: Markus Goeker, with assistance by Pia Wuest

library(BiodiversityR)

## Loading required package: tcltk
## Loading required package: vegan
## Loading required package: permute
## 
## Attaching package: 'permute'
## 
## The following object is masked from 'package:pkgutils':
## 
##     check
## 
## Loading required package: lattice
## 
## Attaching package: 'lattice'
## 
## The following object is masked from 'package:opm':
## 
##     parallelplot
## 
## This is vegan 2.3-0
## BiodiversityR 2.5-2: use function 'BiodiversityRGUI()' to launch the BiodiversityR Graphical User Interface

library(opm)

Define auxiliary function

PCA bi-plot of PM data with BiodiversityR

This function is intended to operate on numeric matrices as generated by opm::extract. It uses methods from the vegan and BiodiversityR packages.

• x Matrix as exported by opm::extract.
• group Logical scalar. Assign symbols by group? This causes secondary symbols to be plotted over the primary symbols (which should best be invisible in that case).
• circle Logical scalar. Draw equilibrium circle?
• text.col Used for drawing descriptor names.
• points.col Used for drawing the primary object symbols.
• arrow.col Used for drawing descriptor arrows.
• legend.x Passed to graphics::legend in non-interactive use.
• legend.y Likewise.
• lwd Used for the secondary object symbols.
• cex Used for drawing descriptor names.
• scaling Passed to stats::biplot.
• ... Optional arguments passed to stats::biplot.

The function invisibly returns an object of class pca.

custom_pca <- function(x, group = TRUE, circle = scaling == 1,
    text.col = "darkgrey", points.col = if (group)
      "white"
    else
      "black", arrow.col = "red", legend.x = "topleft", legend.y = NULL,
    lwd = 2, cex = 0.4, scaling = 1, ...) {
  result <- vegan::rda(x) # conducts a PCA in this setting
  result.plot <- biplot(x = result, col = c(points.col, arrow.col),
    scaling = scaling, ...)
  if (group) {
    groups <- data.frame(.N = if (is.null(attr(x, "row.groups")))
        rownames(x)
      else
        attr(x, "row.groups"))
    if (can.click <- dev.capabilities()$locator)
      message("click on the preferred position of the legend")
    # Problem: in interactive use the returned colour names have been incorrect!
    # This did not seem to work on an X11 device! PDF was OK but see below.
    cols.used <- BiodiversityR::ordisymbol(result.plot, groups, ".N",
      legend = can.click, lwd = lwd)
    if (!can.click)
      legend(legend.x, legend.y, legend = levels(groups$.N),
        text.col = cols.used)
  }
  text(result.plot, "species", col = text.col, cex = cex)
  # Another problem: the circle does not appear in the PDF! Only X11 worked.
  if (circle)
    BiodiversityR::ordiequilibriumcircle(result, result.plot)
  invisible(result)
}

Create PCA bi-plot for the vaas_4 example data

x <- opm::extract(vaas_4, as.labels = list("Strain"),
  as.groups = list("Species"))
x.pca <- custom_pca(x)

detach("package:BiodiversityR", unload = TRUE)
detach("package:vegan", unload = TRUE)