1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
|
% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/clusterSNNGraph.R
\name{clusterSNNGraph}
\alias{clusterSNNGraph}
\alias{clusterSNNGraph,ANY-method}
\alias{clusterSNNGraph,SummarizedExperiment-method}
\alias{clusterSNNGraph,SingleCellExperiment-method}
\alias{clusterKNNGraph}
\alias{clusterKNNGraph,ANY-method}
\alias{clusterKNNGraph,SummarizedExperiment-method}
\alias{clusterKNNGraph,SingleCellExperiment-method}
\title{Wrappers for graph-based clustering}
\usage{
clusterSNNGraph(x, ...)
\S4method{clusterSNNGraph}{ANY}(
x,
...,
clusterFUN = cluster_walktrap,
subset.row = NULL,
transposed = FALSE,
use.kmeans = FALSE,
kmeans.centers = NULL,
kmeans.args = list(),
full.stats = FALSE
)
\S4method{clusterSNNGraph}{SummarizedExperiment}(x, ..., assay.type = "logcounts")
\S4method{clusterSNNGraph}{SingleCellExperiment}(x, ..., use.dimred = NULL)
clusterKNNGraph(x, ...)
\S4method{clusterKNNGraph}{ANY}(
x,
...,
clusterFUN = cluster_walktrap,
subset.row = NULL,
transposed = FALSE,
use.kmeans = FALSE,
kmeans.centers = NULL,
kmeans.args = list(),
full.stats = FALSE
)
\S4method{clusterKNNGraph}{SummarizedExperiment}(x, ..., assay.type = "logcounts")
\S4method{clusterKNNGraph}{SingleCellExperiment}(x, ..., use.dimred = NULL)
}
\arguments{
\item{x}{A matrix-like object containing expression values for each gene (row) in each cell (column).
These dimensions can be transposed if \code{transposed=TRUE}.
Alternatively, a \linkS4class{SummarizedExperiment} or \linkS4class{SingleCellExperiment} containing such an expression matrix.
If \code{x} is a SingleCellExperiment and \code{use.dimred} is set, its \code{\link{reducedDims}} will be used instead.}
\item{...}{For the generics, additional arguments to pass to the specific methods.
For the ANY methods, additional arguments to pass to \code{\link{buildSNNGraph}} or \code{\link{buildKNNGraph}}.
For the SummarizedExperiment method, additional arguments to pass to the corresponding ANY method.
For the SingleCellExperiment method, additional arguments to pass to the corresponding SummarizedExperiment method.}
\item{clusterFUN}{Function that can take a \link{graph} object and return a \link{communities} object,
see examples in the \pkg{igraph} package.}
\item{subset.row}{See \code{?"\link{scran-gene-selection}"}.
Only used when \code{transposed=FALSE}.}
\item{transposed}{A logical scalar indicating whether \code{x} is transposed (i.e., rows are cells).}
\item{use.kmeans}{Logical scalar indicating whether k-means clustering should be performed.}
\item{kmeans.centers}{Integer scalar specifying the number of clusters to use for k-means clustering.
Defaults to the square root of the number of cells in \code{x}.}
\item{kmeans.args}{List containing additional named arguments to pass to \code{\link{kmeans}}.}
\item{full.stats}{Logical scalar indicating whether to return more statistics regarding the k-means clustering.}
\item{assay.type}{A string specifying which assay values to use.}
\item{use.dimred}{A string specifying whether existing values in \code{reducedDims(x)} should be used.}
}
\value{
If \code{full.stats=FALSE}, a factor is returned containing the cluster assignment for each cell.
If \code{full.stats=TRUE} and \code{use.kmeans=TRUE}, a \linkS4class{DataFrame} is returned with one row per cell.
This contains the columns \code{kmeans}, specifying the assignment of each cell to a k-means cluster;
and \code{igraph}, specifying the assignment of each cell to a graph-based cluster operating on the k-means clusters.
In addition, the \code{\link{metadata}} contains \code{graph}, a \link{graph} object where each node is a k-means cluster;
and \code{membership}, the graph-based cluster to which each node is assigned.
}
\description{
Perform graph-based clustering using community detection methods on a nearest-neighbor graph,
where nodes represent cells or k-means centroids.
This has been deprecated in favor of directly using \code{\link{clusterRows}} from the \pkg{bluster} package,
with \code{BLUSPARAM} set to \code{\link{NNGraphParam}()} or \code{\link{TwoStepParam}()}.
}
\details{
We suggest using the \code{\link{clusterRows}} functionality instead as it provides a more general interface to clustering.
\itemize{
\item \code{clusterSNNGraph(x)} for a matrix-like object \code{x} is equivalent to
\code{clusterRows(t(x), NNGraphParam())}.
\item \code{clusterKNNGraph(x)} for a matrix-like object \code{x} is equivalent to
\code{clusterRows(t(x), NNGraphParam(shared=FALSE))}.
\item \code{clusterSNNGraph(x)} for a SummarizedExperiment object \code{x} is equivalent to
\code{clusterRows(t(logcounts(x)), NNGraphParam())}.
\item \code{clusterSNNGraph(x, use.dimred="PCA")} for a SingleCellExperiment object \code{x} is equivalent to
\code{clusterRows(reducedDim(x, "PCA"), NNGraphParam())}.
\item \code{clusterSNNGraph(x, use.kmeans=TRUE, use.dimred="PCA")} for a SingleCellExperiment object \code{x} is equivalent to
\code{clusterRows(reducedDim(x, "PCA"), TwoStepParam())}.
}
}
\examples{
library(scuttle)
sce <- mockSCE(ncells=500)
sce <- logNormCounts(sce)
clusters <- clusterSNNGraph(sce)
table(clusters)
# Can pass usual arguments to buildSNNGraph:
clusters2 <- clusterSNNGraph(sce, k=5)
table(clusters2)
# Works with low-dimensional inputs:
sce <- scater::runPCA(sce, ncomponents=10)
clusters3 <- clusterSNNGraph(sce, use.dimred="PCA")
table(clusters3)
# Turn on k-means for larger datasets, e.g.,
# assuming we already have a PCA result:
set.seed(101010)
bigpc <- matrix(rnorm(2000000), ncol=20)
clusters4 <- clusterSNNGraph(bigpc, d=NA, use.kmeans=TRUE, transposed=TRUE)
table(clusters4)
# Extract the graph for more details:
clusters5 <- clusterSNNGraph(sce, use.dimred="PCA",
use.kmeans=TRUE, full.stats=TRUE)
head(clusters5)
metadata(clusters5)$graph
}
\seealso{
\code{\link{clusterRows}} with \code{BLUSPARAM} set to an instance of \linkS4class{NNGraphParam} or \linkS4class{TwoStepParam}.
}
\author{
Aaron Lun
}
|
