model.test: Copy number model quality assessment
In cghRA: Array CGH Data Analysis and Visualization

Description Usage Arguments Value Author(s) See Also Examples

This function provides various data to manually fit or upgrade a copy number model, as needed by model.apply to translate logRatios into copy numbers.

  model.test(segLogRatios, segChroms, segLengths = rep(1, length(segLogRatios)),
    model = NA, center = model['center'], width = model['width'],
    ploidy = model['ploidy'],  bw = model['bw'], minDensity = 0.001,
    peakFrom = model['peakFrom'], peakTo = model['peakTo'], graph = TRUE,
    parameters = TRUE, returnPar = FALSE, xlim = c(0, 5), ylim = c(0, max(segLengths)),
    xlab = "Segment copy number", ylab = "Segment length", cex.seg = 0.4, cex.leg = 0.7,
    cex.l2r = 0.7, exclude = c("X", "Y", "Xp", "Xq", "Yp", "Yq"), title = NULL,
    panel = FALSE, klim = NULL, ...)

`segLogRatios`	Double vector, the log ratios of the CGH segments to modelize.
`segChroms`	Vector, the chromosome holding the CGH segments to modelize.
`segLengths`	Double vector, the lengths of the CGH segments to modelize. Amount of probes should be prefered if available, but nucleotide length or no length at all can also be used.
`model`	A double vector, as returned by `model.auto` or `model.test`. Can be `NA` if parameters are provided via other arguments.
`center`	Single double value, the center parameter to use in the model.
`width`	Single double value, the width parameter to use in the model.
`ploidy`	Single numeric value, copy number supposed to be the most common within the analyzed genome.
`bw`	Single double value, the bandwidth parameter to use in the model.
`minDensity`	Single double value, minimal density for a peak to be detected.
`peakFrom`	Single double value, the peak logRatio lower limit parameter to use in the model.
`peakTo`	Single double value, the peak logRatio upper limit parameter to use in the model.
`graph`	Single logical value, whether to plot the density distribution of the segments with the modelized copy numbers or not.
`parameters`	Single logical value, whether to add a legend to the plot with the parameters and statistics of the model or not.
`returnPar`	Single logical value, whether to return the `par` content (for point identification in interactive plots) or the model statistics.
`xlim`	Vector of two double values, the boundaries of the plot on the horizontal axis (in `LCN`).
`ylim`	Vector of two double values, the boundaries of the plot on the vertical axis (in the same units than `segLengths`).
`xlab`	Single character value, the title to print for the horizontal axis.
`ylab`	Single character value, the title to print for the vertical axis.
`cex.seg`	Single double value, the character expansion factor for points (segments) on the plot.
`cex.leg`	Single double value, the character expansion factor for the plot legend.
`cex.l2r`	Single double value, the character expansion factor for the log-ratio axis of the plot.
`exclude`	Vector, the chromosomes to exclude from the density computation and to plot with distinct symbols (use `NULL` to disable this feature). Sexual chromosomes should be excluded in heterogeneous DNA source, as their desequilibrium (2 'X' and no 'Y' in women) impact normal cells AND tumoral ones.
`title`	To be passed to `legend`, see there for allowed types (usually a single character value).
`panel`	Single logical value, whether to plot a rotated minimalist graph or a classic one.
`klim`	Double vector of two values, alternative definition of `xlim` in modelized copy numbers rather than LCN.
`...`	Further graphical arguments to be passed to `plot`.

When returnPar is TRUE, invisibly returns the par content, for point identification.

When returnPar is FALSE, returns the same vector as model.auto, see its help page for further details.

Sylvain Mareschal

model.auto, model.apply

  # Generating random segmentation results
  ## with 30% normal cells contamination
  ## with +10% for normal DNA labelling
  segLogRatios <- c(
    rnorm(
      sample(5:20, 1),
      mean = log((1*0.7 + 2*0.3)/(2*1.1), 2),   # One deletion
      sd = 0.08
    ),
    rnorm(
      sample(80:120, 1),
      mean = log(2/(2*1.1), 2),                 # No alteration
      sd = 0.08
    ),
    rnorm(
      sample(40:60, 1),
      mean = log((3*0.7 + 2*0.3)/(2*1.1), 2),   # One more copy
      sd = 0.08
    )
  )
  segLogRatios <- sample(segLogRatios)
  segLengths <- as.integer(3 + round(rchisq(length(segLogRatios), 1)*100))
  segEnds <- cumsum(segLengths)
  segStarts <- c(1L, head(segEnds, -1))
  segChroms <- rep("chr1", length(segEnds))
  
  # Generated genome
  genome <- data.frame(
    segChroms,
    segStarts,
    segEnds,
    segLogRatios,
    segLengths
  )
  print(genome)
  
  # Automatic modelization
  autoModel <- model.auto(
    segLogRatios = segLogRatios,
    segChroms = segChroms,
    segLengths = segLengths
  )
  
  layout(matrix(1:2, ncol=1))
  
  # Show automatic model
  model.test(
    segLogRatios = segLogRatios,
    segChroms = segChroms,
    segLengths = segLengths,
	model = autoModel
  )
  
  # Standard model derived from the log ratios definition
  refModel <- model.test(
    segLogRatios = segLogRatios,
    segChroms = segChroms,
    segLengths = segLengths,
    center = 2,
    width = 1,
    bw = 0.1        # Arbitrary
  )
  
  # Differences in scores
  print(autoModel)
  print(refModel)
  
  layout(1)