MyClustering: Semi-Supervised Clustering Algorithm

In dsherma7/PolyploidDosageCalling: Calls the Dosage From Tetraploid qPCR Data (CT/RQ Values)

Description

Following the standard K-Means algorithm, we start by assigning each sample a random cluster value. Howewver, the known control values are assigned their own dosage. Then this algorithm performs consistent iterations of computing the mean of each cluster and then relabeling each sample to it's closest cluster.

Usage

MyClustering(data, nClust = 5, Labels = rep(NA, nrow(X)), max.iter = 1000,
  converge.threshold = 0.05, outlier.threshold = 2, get.Prob = TRUE)

Arguments

data	Dataset used for clustering. Must be an n x p numeric matrix; where n is the number of samples and p is the number of features.
nClust	Expected total number of clusters (i.e the possible dosage values). Default is 5 for tetraploids.
Labels	The initial labels of all the data. Unknown samples are marked NA, while the controls are dosage +1 (since 0 is invalid).
max.iter	Max number of iterations allowed to run for this algorithm. If set too low a warning will occur.
converge.threshold	If the percent change between two iterations is within this value then the algorithm will stop. Makes this more effecient.
outlier.threshold	Used to determine the probability values. Samples outside with standard deviation larger than this value will have very low probability and will be thrown out in determining the remaining samples' probability.
get.Prob	TRUE/FALSE value indicating if probabilities are desired alongside dosage calls. Defaults to TRUE, but use FALSE for more efficiency if probability is not needed.f

dsherma7/PolyploidDosageCalling documentation built on May 23, 2019, 6:06 p.m.