APJaccardSimKRange: A function which finds the clustering solutions for cells in...

View source: R/APJaccard.R

APJaccardSimKRangeR Documentation

A function which finds the clustering solutions for cells in a Seurat object based on transformed Jaccard similarity matrices with different numbers of nearest neighbors. Can also return nmis for these solutions and the similarity matrices from which they were calculated

Description

A function which finds the clustering solutions for cells in a Seurat object based on transformed Jaccard similarity matrices with different numbers of nearest neighbors. Can also return nmis for these solutions and the similarity matrices from which they were calculated

Usage

APJaccardSimKRange(
  data.dist,
  transformation = "default",
  APq = 0,
  kmin = 5,
  kmax = 50,
  kstep = 5,
  target.clust.vect = NULL,
  save.simmats = FALSE
)

Arguments

data.dist

a distance matrix on which to calculate nearest neighbors

transformation

a character vector indicating the type of transformation the similarity matrix will go through before being returned. By default the similarity matrix will be returned with no transformation. Other possible values include log (returns matrix which has been log transformed), 'inverse' returns a shifted inverse of the matrix, 'stretch' returns a linear scaling of the matrix, 'tan' returns the tangent of matrix values, and 'tanh' returns the hyperbolic tangent of matrix values

APq

a number between 0 and 1 which indicates the quantile of the data to use to diagonalize the similarity matrix before running AP clustering. Increase if too few clusters.

kmin

the number of neighbors on which to calculate the first Jaccard matrix

kmax

the number of neighbors on which to calculate the last Jaccard matrix

kstep

the number of neighbors to iterate for each calculation of a new Jaccard matrix

save.simmats

a Boolean indicating whether or not to return the Jaccard similarity matrices along with the AP clustering solutions.

target.clust.vec

a named vector of cluster assignments for the cells/samples in the distance matrix. If given, the function returns the NMI values for the AP solutions for each Jaccard matrix when compared with this target vector. If null, the function returns no NMI values

Value

a list containing the AP clustering solutions for all the variations of the Jaccard matrix and, if there is a target vector given, the NMI values for these solutions compared with the target vector and, if save.simmat=TRUE, the similarity matrix which resulted in each AP solution


arh2207/APJaccard documentation built on May 19, 2022, 11:59 p.m.