View source: R/benchmark.clustering.R
benchmark.clustering | R Documentation |
Supervised (ARI and NMI) and unsupervised (ASW, Dunn Index, and Connectivity) benchmarking metrics are calculated for cluster assignments. assays, clustering and reduction for distance calculations are iterated through.
benchmark.clustering( object, assay, clustering, reduction, n.dims = 2, dist.method = "euclidean", ground.truth.column = NULL, threads = 1, verbose = FALSE, seed = 1234 )
object |
IBRAP S4 class object |
assay |
Character. String containing indicating which assay to use |
clustering |
Character. The names of the cluster assignment dataframes to use |
reduction |
Character. Which reduction(s) within the assay should be supplied for distance calcultions |
n.dims |
Numerical. How many dimensions of the reduction should be supplied. Default = 1:3 |
dist.method |
Character. Which distance method should be used, options: 'euclidean', 'maximum', 'manhattan', 'canberra', 'binary', 'minkowski'. Default = 'euclidean' |
ground.truth.column |
Character. If available, supply the column in the object metadata that contains ground truth labels, i.e. true cell type labels. If this is not supplied, only unsupervised methods will be supplied. Default = NULL |
verbose |
Logical. Should system information be printed. Default = FALSE |
seed |
Numeric. What should the seed be set as. Default = 1234 |
Benchmarking scores for the supplied cluster assignments
# without ground truth labels object <- benchmark.clustering(object = object, assay = c('SCT', 'SCRAN', 'SCANPY'), clustering = c("pca_harmony_nn:louvain"), reduction = c('pca_harmony_umap'), n.dims = 1:2) # With ground truth labels object <- benchmark.clustering(object = object, assay = c('SCT', 'SCRAN', 'SCANPY'), clustering = c("pca_harmony_nn.v1:louvain"), reduction = c('pca_harmony_umap'), n.dims = 1:2, ground.truth = metadata$celltypes)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.