groupSamples: Performs community detection on sample-sample distance matrix...

In wschen/phemd: Phenotypic EMD for comparison of single-cell samples

Performs community detection on sample-sample distance matrix to identify groups of similar samples

Description

Takes sample-sample distance matrix as input and returns group assignments for each sample

Usage

groupSamples(
  distmat,
  distfun = "hclust",
  ncluster = NULL,
  method = "complete",
  ...
)

Arguments

distmat	A distance matrix of dimension num_samples x num_samples representing pairwise dissimilarity between samples
distfun	Method of partitioning network of samples (currently either 'hclust' or 'pam')
ncluster	Optional parameter specifying total number of sample groups
method	Optional parameter for hierarchical clustering (see "hclust" documentation)
...	Optional additional parameters to be passed to diffusionKmeans method

Details

By default, uses 'kgs' (Kelley-Gardner-Sutcliffe) method for determining optimal number of groups. Alternatively, can take user-specified number of groups). Requires 'cluster' and 'maptree' packages.

Value

Vector containing group assignments for each sample (same order as row-order of distmat) based on user-specified partitioning method (e.g. hierarchical clustering)

Examples

my_phemdObj <- createDataObj(all_expn_data, all_genes, as.character(snames_data))
my_phemdObj_lg <- removeTinySamples(my_phemdObj, 10)
my_phemdObj_lg <- aggregateSamples(my_phemdObj_lg, max_cells=1000)
my_phemdObj_monocle <- embedCells(my_phemdObj_lg, cell_model = 'monocle2', data_model = 'gaussianff', sigma=0.02, maxIter=2)
my_phemdObj_monocle <- orderCellsMonocle(my_phemdObj_monocle)
my_phemdObj_final <- clusterIndividualSamples(my_phemdObj_monocle)
my_phemdObj_final <- generateGDM(my_phemdObj_final)
my_EMD_mat <- compareSamples(my_phemdObj_final)
cluster_assignments <- groupSamples(my_EMD_mat, distfun = 'hclust', ncluster=4)

wschen/phemd documentation built on Oct. 1, 2024, 4:45 a.m.