processingCluster {flowSpy} | R Documentation |
Calculate Principal Components Analysis (PCA), t-Distributed Stochastic Neighbor Embedding (tSNE), Diffusion Map and Uniform Manifold Approximation and Projection (UMAP) of clusters calculated by runCluster.
processingCluster( object, perplexity = 5, k = 5, downsampling.size = 1, force.resample = TRUE, random.cluster = FALSE, umap.config = umap.defaults, verbose = FALSE, ... )
object |
an FSPY object |
perplexity |
numeric. Perplexity parameter (should not be bigger than 3 *
perplexity < nrow(X) - 1, see details for interpretation). See |
k |
numeric. The parameter k in k-Nearest Neighbor. |
downsampling.size |
numeric. Percentage of sample size of downsampling. This parameter is from 0 to 1. by default is 1. |
force.resample |
logical. Whether to do resample if downsampling.size < 1 |
random.cluster |
logical. Whether to perfrom random downsampling. If FALSE, an uniform downsampling will be processed. |
umap.config |
object of class umap.config. See |
verbose |
logic. Whether to print calculation progress. |
... |
options to pass on to the dimensionality reduction functions. |
An FSPY object with cluster.id in meta.data
An FSPY object with dimensionality reduction of clusters
umap
, fast.prcomp
,
Rtsne
, destiny
if (FALSE) { # After running clustering set.seed(1) fspy <- runCluster(fspy, cluster.method = "som", xdim = 3, ydim = 3, verbose = T) # Do not perfrom downsampling fspy <- processingCluster(fspy, perplexity = 2) # Perform cluster based downsampling # Only keep 50% cells fspy <- processingCluster(fspy, perplexity = 2, downsampling.size = 0.5) # Processing clusters without downsampling step fspy <- processingCluster(fspy, perplexity = 2, force.resample = FALSE) }