Run Non-Negative Matrix Factorization

Use NMF to perform dimension reduction.

runNMF(
  gobject,
  spat_unit = NULL,
  feat_type = NULL,
  expression_values = c("normalized", "scaled", "custom"),
  reduction = c("cells", "feats"),
  name = NULL,
  feats_to_use = "hvf",
  return_gobject = TRUE,
  scale_unit = TRUE,
  k = 20,
  method = c("rcppml"),
  rev = FALSE,
  set_seed = TRUE,
  seed_number = 1234,
  verbose = TRUE,
  toplevel = 1L,
  ...
)

Arguments

gobject: giotto object
spat_unit: spatial unit (e.g. "cell")
feat_type: feature type (e.g. "rna", "dna", "protein")
expression_values: expression values to use
reduction: "cells" or "feats"
name: arbitrary name for NMF run
feats_to_use: subset of features to use for NMF
return_gobject: boolean: return giotto object (default = TRUE)
scale_unit: scale features before NMF (default = TRUE)
k: NMF rank (number of components to decompose into). Default is 20
method: which implementation to use (only rcppml right now)
rev: do a reverse NMF
set_seed: use of seed
seed_number: seed number to use
verbose: verbosity of the function
toplevel: relative stackframe where call was made
...: additional parameters for NMF (see details)

Value

giotto object with updated NMF dimension reduction

Details

See nmf for more information about other parameters.

Examples

g <- GiottoData::loadGiottoMini("visium")
#> 1. read Giotto object
#> 2. read Giotto feature information
#> 3. read Giotto spatial information
#> 3.1 read Giotto spatial shape information
#> 3.2 read Giotto spatial centroid information
#> 3.3 read Giotto spatial overlap information
#> 4. read Giotto image information
#> python already initialized in this session
#>  active environment : '/usr/bin/python3'
#>  python version : 3.12
x <- runNMF(g, k = 20)
#> "hvf" column was found in the feats metadata information and will be
#>  used to select highly variable features
#> [Running RcppML NMF]. This citation shown once per session:
#> Zachary J. DeBruine, Karsten Melcher, Timothy J. Triche Jr
#>  Fast and robust non-negative matrix factorization for single-cell experiments
#>  bioRxiv 2021.09.01.458620.
#>  https://doi.org/10.1101/2021.09.01.458620
#> 
#> iter |      tol 
#> ---------------
#>    1 | 8.68e-01
#>    2 | 1.06e-01
#>    3 | 4.20e-02
#>    4 | 2.40e-02
#>    5 | 1.50e-02
#>    6 | 1.07e-02
#>    7 | 8.13e-03
#>    8 | 5.39e-03
#>    9 | 3.11e-03
#>   10 | 2.04e-03
#>   11 | 1.53e-03
#>   12 | 1.18e-03
#>   13 | 9.46e-04
#>   14 | 7.65e-04
#>   15 | 6.38e-04
#>   16 | 5.10e-04
#>   17 | 4.68e-04
#>   18 | 4.30e-04
#>   19 | 3.94e-04
#>   20 | 3.51e-04
#>   21 | 3.07e-04
#>   22 | 2.39e-04
#>   23 | 1.90e-04
#>   24 | 1.70e-04
#>   25 | 1.53e-04
#>   26 | 1.35e-04
#>   27 | 1.21e-04
#>   28 | 1.05e-04
#>   29 | 9.11e-05
#> Setting dimension reduction [cell][rna] nmf
x <- runUMAP(x,
    dim_reduction_to_use = "nmf",
    dimensions_to_use = 1:20,
    name = "nmf_umap"
)
#> Setting dimension reduction [cell][rna] nmf_umap
x <- createNearestNetwork(x,
    dim_reduction_to_use = "nmf",
    dim_reduction_name = "nmf",
    dimensions_to_use = 1:20
)
x <- doLeidenCluster(x, name = "nmf_leiden", network_name = "sNN.nmf")
#> Error in py_run_file_impl(file, local, convert): ModuleNotFoundError: No module named 'igraph'
#> Run `reticulate::py_last_error()` for details.
plotUMAP(x, dim_reduction_name = "nmf_umap", cell_color = "nmf_leiden")
#> Error: dimPlot2D()
#>  nmf_leiden is not a color or a column name
spatPlot2D(x, cell_color = "nmf_leiden")
#> Error: spatPlot2D()
#>  nmf_leiden is not a color or a column name