Skip to contents

Convert an SNN graph to a dense similarity matrix

Source: R/clusterMethods.R
snnToMatrix.Rd

Helper for writing clustering methods that operate on a full similarity matrix rather than a graph (affinity propagation, binary cut, k-means on similarity profiles). The SNN graph is already a similarity matrix with values in [0, 1]; this simply densifies it and keeps the gene set names as dimnames.

Usage

snnToMatrix(snn, warnAbove = 5000)

Arguments

snn

The SNN graph passed to a cluster method.

warnAbove

Number of gene sets above which a warning about the memory cost of densifying is emitted. Set to Inf to silence. The dense matrix needs roughly 8 * N^2 bytes, so 10,000 gene sets is about 800 MB.

Value

A dense symmetric numeric matrix with gene sets as rows/columns.

See also

clusterMethod(), snnToIgraph()

Examples

simMat <- snnToMatrix(exampleGsList@cluster@graphs$GSEA_snn)
simMat[1:3, 1:3]
#>                                           GRADE_COLON_AND_RECTAL_CANCER_UP
#> GRADE_COLON_AND_RECTAL_CANCER_UP                                         1
#> BENPORATH_NANOG_TARGETS                                                  0
#> GOBERT_OLIGODENDROCYTE_DIFFERENTIATION_UP                                0
#>                                           BENPORATH_NANOG_TARGETS
#> GRADE_COLON_AND_RECTAL_CANCER_UP                                0
#> BENPORATH_NANOG_TARGETS                                         1
#> GOBERT_OLIGODENDROCYTE_DIFFERENTIATION_UP                       0
#>                                           GOBERT_OLIGODENDROCYTE_DIFFERENTIATION_UP
#> GRADE_COLON_AND_RECTAL_CANCER_UP                                                  0
#> BENPORATH_NANOG_TARGETS                                                           0
#> GOBERT_OLIGODENDROCYTE_DIFFERENTIATION_UP                                         1