knnClassify: Use the k-nearest neighbour algorithm to classify the wells...

In twoddpcr: Classify 2-d Droplet Digital PCR (ddPCR) data and quantify the number of starting molecules

Description

If droplets is a data frame, the droplets are classified using the k-nearest neighbour algorithm against a training data set.

If droplets is a ddpcrWell object, the droplets in the well are classified and returned in another ddpcrWell object.

If droplets is a ddpcrPlate object, the wells are combined and classified together, with the resulting classification assigned to the ddpcrPlate object.

Usage

knnClassify(droplets, trainData, cl, k, prob = 0, ...)

## S4 method for signature 'data.frame'
knnClassify(droplets, trainData, cl, k, prob = 0, fullTable = TRUE)

## S4 method for signature 'ddpcrWell'
knnClassify(droplets, trainData, cl, k, prob = 0)

## S4 method for signature 'ddpcrPlate'
knnClassify(droplets, trainData, cl, k, prob = 0)

Arguments

droplets	A ddpcrWell or ddpcrPlate object, or a data frame with columns Ch1.Amplitude and Ch2.Amplitude.
trainData	A data frame of training data with columns Ch1.Amplitude and Ch2.Amplitude.
cl	A vector of classes corresponding to trainData.
k	The number of nearest neighbours to use in the algorithm.
prob	The minimal proportion of votes for the winning class needed to assert that a droplet belongs to the class. This figure should be a float between 0 and 1. For example, if 0.6 then at least 60 k-nearest neighbours need to be of one class, otherwise it is classified as "Rain". Defaults to 0, i.e. we do not use "Rain".
...	Other options depending on the type of droplets.
fullTable	If TRUE, returns a full data frame of droplets and their classification; if FALSE, simply returns the classified vector. Defaults to TRUE.

Value

An object with the new classification.

If droplets is a data frame, return data frame or factor (depending on the value of fullTable) of the droplet classification under the k-NN algorithm.

Author(s)

Anthony Chiu, [e-mail]

See Also

This method uses the knn function.

To manually set and retrieve classifications, use the wellClassification, plateClassification and plateClassificationMethod methods.

kmeansClassify provides a wrapper for the k-means clustering algorithm.

Examples

### Use the KRASdata dataset for all of these examples.

## Use k-means clustering to classify one sample. Use this as training
## data for the K-Nearest Neighbour algorithm.
trainingData <- KRASdata[["E03"]]
trainingData <- kmeansClassify(trainingData)$data

## Classify a dataframe using k-NN with k = 1 and the above training data.
aWell <- knnClassify(
           KRASdata[["F03"]],
           trainData=trainingData[, c("Ch1.Amplitude", "Ch2.Amplitude")],
           cl=trainingData$class,
           k=1)
dropletPlot(aWell, cMethod="class")  # visualising the classification

## We can change k to a larger number, here with a ddpcrWell object.
aWell <- ddpcrWell(well=KRASdata[["E03"]])
aWell <- knnClassify(
           aWell,
           trainData=trainingData[, c("Ch1.Amplitude", "Ch2.Amplitude")],
           cl=trainingData$class,
           k=3)
dropletPlot(aWell, cMethod="knn")  # visualising the classification

## Changing the 'prob' parameter means that droplets with less than 'prob'
## of the votes will not be classified. We do this for a ddpcrPlate
## object.
krasPlate <- ddpcrPlate(wells=KRASdata[c("E03", "H03", "C04", "F04")])
krasPlate <- knnClassify(
               krasPlate,
               trainData=trainingData[, c("Ch1.Amplitude", "Ch2.Amplitude")],
               cl=trainingData$class,
               k=3,
           prob=0.6)
dropletPlot(krasPlate, cMethod="knn")  # visualising the classification

twoddpcr documentation built on Nov. 8, 2020, 5:49 p.m.