RunNMF: Run Non-negative Matrix Factorization

In jbergenstrahle/STUtility: Analysis and visualization of Spatial Transcriptomics data

Run Non-negative Matrix Factorization

Description

Decompose an expression matrix A with non-negative elements into matrices WxH, also with non-negative elements. W is the feature loading matrix (features x factors) and H is the low dimensional embedding of the spots (factors x spots).

Usage

RunNMF(
  object,
  assay = NULL,
  slot = "scale.data",
  features = NULL,
  nfactors = 20,
  rescale = TRUE,
  reduction.name = "NMF",
  reduction.key = "factor_",
  n.cores = NULL,
  order.by.spcor = FALSE,
  sort.spcor.by.var = FALSE,
  ...
)

Arguments

object	Seurat object
assay	Assay Name of Assay NMF is being run on
slot	Slot to pull data from.
features	Features to compute the NMF for. Note that these features must be present in the slot used to compute the NMF. By default, the 'features' is set to 'VariableFeatures(object)' to include the most variable features selected in the normalization step.
nfactors	Total Number of factors to compute and store (20 by default)
rescale	Rescale data to make sure that values of the input matrix are non-n
reduction.name	Dimensional reduction name, "NMF" by default
reduction.key	Dimensional reduction key, specifies the prefix of the factor ids, e.g. "factor_1", "factor_2", etc.
n.cores	Number of threads to use in computation
order.by.spcor	Order factors by spatial correlation
sort.spcor.by.var	Sort factors by decreasing variance
...	Additional parameters

jbergenstrahle/STUtility documentation built on March 14, 2023, 7:15 a.m.