renormalisePlate: Normalise all wells in a plate to the overall/average cluster...

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

Description

Each well is scaled in the direction of the two axes independently. The algorithm works as follows:

1. Classify the entire plate using k-means clustering in order to roughly identify clusters.

2. For each channel and each well:

   1. Remove the positive clusters in the other channel.

   2. Re-run k-means clustering with k = 2 to obtain new cluster centres. If the centres are too close (default distance 2000), reject these new cluster centres and use the old centres.

   3. Use the new centres for each well to rescale to the same scale as the average/overall scale.

Usage

renormalisePlate(
  plate,
  initialCentres = matrix(c(0, 0, 10000, 0, 0, 7000, 10000, 7000), ncol = 2, byrow =
    TRUE),
  minSeparation = 2000
)

Arguments

plate	A ddpcrPlate object.
initialCentres	A matrix or data frame of (rough estimates) of centres of each of the clusters in the combined data. This will be used for an initial run of k-means clustering.
minSeparation	The minimum distance required between two cluster centres in order for us to assume that k-means clustering found two distinct clusters. This is used when classifying droplets along a single channel and helps to reject classifications where there is only one cluster, e.g. when there are no mutants in a well. Defaults to 2000.

Value

A list of data frames with the rescaled amplitudes in both channels.

Author(s)

Anthony Chiu, [e-mail]

Examples

## Normalise the KRAS data.
plate <- ddpcrPlate(wells=KRASdata)
normPlate <- renormalisePlate(plate)

CRUKMI-ComputationalBiology/twoddpcr documentation built on Feb. 14, 2021, 9:18 p.m.