hierBipartite {hierBipartite}

Bipartite Graph-based Hierarchical Clustering

Description

Main bipartite graph-based hierarchial clustering algorithm. Visit here for vignette on using the hierBipartite package.

Usage

hierBipartite(
  X,
  Y,
  groups,
  link = "ward.D2",
  n_subsample = 1,
  subsampling_ratio = 1,
  p.value = FALSE,
  n_perm = 100,
  parallel = FALSE,
  maxCores = 7,
  p_cutoff = 0.1
)

Arguments

X

an n x p matrix of variable set 1 (e.g. gene expression)

Y

an n x q matrix of variable set 2 (e.g. drug sensitivity)

groups

a list of starting group membership (e.g. list("1" = c(1,2,3), "2" = c(4,5,6)) means group 1 has samples 1, 2, 3, and group 2 has samples 4, 5, 6.

link

string indicating link function as input to hclust(). One of "ward.D", "ward.D2", "single", "complete", "average", "mcquitty", "median", "centroid".

n_subsample

number of subsampling to generate matrix B (see paper)

subsampling_ratio

fraction of samples to sample for subsampling to generate matrix B (see paper)

p.value

boolean for whether to generate p-values for each merge

n_perm

number of permutations for generating p-values. Ignored if p.value = FALSE

parallel

boolean for whether to parallelize subsampling and p-value generation step

maxCores

maximum number of cores to use (only applicable when parallel = TRUE)

p_cutoff

p-value cutoff that determines whether merge is significant. If p-value > p_cutoff, p-values will not be calculated for future merges involving current group. Ignored if p.value = FALSE.

Value

list of results from bipartite graph-based hierarchical clustering, containing up to

• hclustObj: hclust object

• groupMerges: list of clusters after each merge, in order of merge. Each cluster is indicated by a vector of cell line groups

• nodePvals: list of p-value of each new merge, in order of merge. Only available if p.value = TRUE

• D: dissimilarity matrix

Examples

# Get a small subset of the test dataset
data(ctrp2)

groups = ctrp2$groups
X = ctrp2$X
Y = ctrp2$Y

groupNames = names(groups)
groupSmall = groups[groupNames[1:3]]

## Not run: 
# Basic call of hierBipartite() on small test dataset
result0 = hierBipartite(X, Y, groupSmall)

# Calling hierBipartite() with subsampling
result1 = hierBipartite(X, Y, groupSmall, n_subsample = 100, subsampling_ratio = 0.90)

# Calling hierBipartite() with p-value generation
result2 = hierBipartite(X, Y, groupSmall, n_perm = 100, p.value = TRUE, p_cutoff = 0.10)

# Calling hierBipartite() with both subsampling and p-value generation (expensive)
result3 = hierBipartite(X, Y, groupSmall, n_subsample = 100, subsampling_ratio = 0.90,
                        n_perm = 100, p.value = TRUE, p_cutoff = 0.10)

## End(Not run)