R: Use a 'grid' to create training data for classification...

gridClassify {twoddpcr}

R Documentation

Use a 'grid' to create training data for classification algorithms.

Description

Classify droplets as "NN", "NP", "PN" or "PP". The classification is based on upper bounds for negative readings and lower bounds for positive readings; see the details and parameters for more detail. If required (see the trainingData parameter), droplets that are not classified will be given the label "N/A".

Usage

gridClassify(droplets, ch1NNThreshold = 6500, ch2NNThreshold = 1900,
  ch1NPThreshold = 6500, ch2NPThreshold = 5000, ch1PNThreshold = 10000,
  ch2PNThreshold = 2900, ch1PPThreshold = 7500, ch2PPThreshold = 5000,
  ...)

## S4 method for signature 'data.frame'
gridClassify(droplets, ch1NNThreshold = 6500,
  ch2NNThreshold = 1900, ch1NPThreshold = 6500, ch2NPThreshold = 5000,
  ch1PNThreshold = 10000, ch2PNThreshold = 2900, ch1PPThreshold = 7500,
  ch2PPThreshold = 5000, trainingData = TRUE, fullTable = TRUE,
  naLabel = ddpcr$rain)

## S4 method for signature 'ddpcrWell'
gridClassify(droplets, ch1NNThreshold = 6500,
  ch2NNThreshold = 1900, ch1NPThreshold = 6500, ch2NPThreshold = 5000,
  ch1PNThreshold = 10000, ch2PNThreshold = 2900, ch1PPThreshold = 7500,
  ch2PPThreshold = 5000, classMethodLabel = "grid", naLabel = ddpcr$rain)

## S4 method for signature 'ddpcrPlate'
gridClassify(droplets, ch1NNThreshold = 6500,
  ch2NNThreshold = 1900, ch1NPThreshold = 6500, ch2NPThreshold = 5000,
  ch1PNThreshold = 10000, ch2PNThreshold = 2900, ch1PPThreshold = 7500,
  ch2PPThreshold = 5000, classMethodLabel = "grid", naLabel = ddpcr$rain)

Arguments

`droplets`	A `ddpcrWell` object or a data frame of droplet amplitudes with columns `Ch1.Amplitude` and `Ch2.Amplitude`.
`ch1NNThreshold`	The channel 1 upper bound for the NN class. Defaults to 6500.
`ch2NNThreshold`	The channel 2 upper bound for the NN class. Defaults to 1900.
`ch1NPThreshold`	The channel 1 upper bound for the NP class. Defaults to 6500.
`ch2NPThreshold`	The channel 2 lower bound for the NP class. Defaults to 5000.
`ch1PNThreshold`	The channel 1 lower bound for the PN class. Defaults to 10000.
`ch2PNThreshold`	The channel 2 upper bound for the PN class. Defaults to 2900.
`ch1PPThreshold`	The channel 1 lower bound for the PP class. Defaults to 7500.
`ch2PPThreshold`	The channel 2 lower bound for the PP class. Defaults to 5000.
`...`	Other options depending on the type of `droplets`.
`trainingData`	Whether to use the output as training data. If `TRUE`, returns the _full table_ with the "N/A" entries removed; if `FALSE`, the "N/A" entries are retained. Taken to be `FALSE` if `fullTable` is set to `FALSE`. Defaults to `TRUE`. Ignored if `droplets` is not a data frame.
`fullTable`	Whether to return a data frame including amplitude figures. If `TRUE`, a data frame with columns `Ch1.Amplitude`, `Ch1.Amplitude` and `class` is returned. If `FALSE`, a factor with levels in `ddpcr$classesRain` is returned, where each entry corresponds to each row in `droplets` (and `trainingData` is automatically set to `FALSE`). Defaults to `TRUE`. Ignored if `droplets` is not a data frame.
`naLabel`	The label to use for unclassified droplets. Should be either ddpcr$na ("N/A") or ddpcr$rain ("Rain"). Defaults to ddpcr$rain.
`classMethodLabel`	A name (as a character string) of the classification method. Defaults to "grid".

Details

The threshold parameters correspond to those in the following diagram:

Ch1 ^        |  |
    |        |  |
    |   PN   |  |
    |        |  |   PP
   e|________|  |
   g|........:..|_________
    |        :  :
   c|.............._______
   a|______  :  : |
    |      | :  : |  NP
    |  NN  | :  : |
    |      | :  : |
    --------------------->
           b f  h d    Ch2

Specifically:

a:: ch1NNThreshold,
b:: ch2NNThreshold,
c:: ch1PNThreshold,
d:: ch2PNThreshold,
e:: ch1NPThreshold,
f:: ch2NPThreshold,
g:: ch1PPThreshold,
h:: ch2PPThreshold.

Value

If droplets is a data frame, return a data frame or factor (depending on the trainingData and fullTable parameters) with a classification for droplets in the chosen regions.

If droplets is a ddpcrWell object, return a ddpcrWell object with the appropriate classification.

If droplets is a ddpcrPlate object, return a ddpcrPlate object with the appropriate classification.

Author(s)

Anthony Chiu, anthony.chiu@cruk.manchester.ac.uk

Examples

## Use a grid to set training data for a data frame.
sgCl <- gridClassify(KRASdata[["E03"]],
                     ch1NNThreshold=5700, ch2NNThreshold=1700,
                     ch1NPThreshold=5400, ch2NPThreshold=5700,
                     ch1PNThreshold=9700, ch2PNThreshold=2050,
                     ch1PPThreshold=7200, ch2PPThreshold=4800)
str(sgCl)

## For data frame only, we can set the trainingData flag to FALSE so that
## the unclassified droplets are retained but labelled as "N/A"
sgCl <- gridClassify(KRASdata[["E03"]],
                     ch1NNThreshold=5700, ch2NNThreshold=1700,
                     ch1NPThreshold=5400, ch2NPThreshold=5700,
                     ch1PNThreshold=9700, ch2PNThreshold=2050,
                     ch1PPThreshold=7200, ch2PPThreshold=4800,
                     trainingData=FALSE)
dropletPlot(sgCl, cMethod="class")

## The same works for ddpcrWell objects.
aWell <- ddpcrWell(well=KRASdata[["E03"]])
aWell <- gridClassify(aWell,
                      ch1NNThreshold=5700, ch2NNThreshold=1700,
                      ch1NPThreshold=5400, ch2NPThreshold=5700,
                      ch1PNThreshold=9700, ch2PNThreshold=2050,
                      ch1PPThreshold=7200, ch2PPThreshold=4800)
str(aWell)

## ddpcrPlate objects work in exactly the same way.
krasPlate <- ddpcrPlate(wells=KRASdata)
krasPlate <- gridClassify(krasPlate)
lapply(plateClassification(krasPlate, withAmplitudes=TRUE), head, n=1)

## The default classification method (column name) is 'gridClassify',
## which may be a bit long. It can be changed.
krasPlate <- gridClassify(krasPlate, classMethodLabel="training")
lapply(plateClassification(krasPlate, withAmplitudes=TRUE), head, n=1)

[Package twoddpcr version 1.11.0 Index]