Run NMF

Run NMF

Usage

RunNMF(object, ...)

# S3 method for class 'Seurat'
RunNMF(
  object,
  assay = NULL,
  layer = "data",
  features = NULL,
  nbes = 50,
  nmf.method = "RcppML",
  tol = 1e-05,
  maxit = 100,
  rev.nmf = FALSE,
  ndims.print = 1:5,
  nfeatures.print = 30,
  reduction.name = "nmf",
  reduction.key = "BE_",
  verbose = TRUE,
  seed.use = 11,
  cores = 0,
  ...
)

# S3 method for class 'Assay'
RunNMF(
  object,
  assay = NULL,
  layer = "data",
  features = NULL,
  nbes = 50,
  nmf.method = "RcppML",
  tol = 1e-05,
  maxit = 100,
  rev.nmf = FALSE,
  ndims.print = 1:5,
  nfeatures.print = 30,
  reduction.key = "BE_",
  verbose = TRUE,
  seed.use = 11,
  cores = 0,
  ...
)

# S3 method for class 'Assay5'
RunNMF(
  object,
  assay = NULL,
  layer = "data",
  features = NULL,
  nbes = 50,
  nmf.method = "RcppML",
  tol = 1e-05,
  maxit = 100,
  rev.nmf = FALSE,
  ndims.print = 1:5,
  nfeatures.print = 30,
  reduction.key = "BE_",
  verbose = TRUE,
  seed.use = 11,
  cores = 0,
  ...
)

# Default S3 method
RunNMF(
  object,
  assay = NULL,
  layer = "data",
  nbes = 50,
  nmf.method = "RcppML",
  tol = 1e-05,
  maxit = 100,
  rev.nmf = FALSE,
  ndims.print = 1:5,
  nfeatures.print = 30,
  reduction.key = "BE_",
  verbose = TRUE,
  cores = 0,
  seed.use = 11,
  ...
)

Arguments

object

An object. This can be a Seurat object, an Assay object, or a matrix-like object.

...

Additional arguments passed to RcppML::nmf or NMF::nmf.

assay

Assay to use. NULL uses the default assay.

layer

Assay layer to use.

features

Features used instead of a reduction.

nbes

The number of basis vectors (components) to be computed.

nmf.method

The NMF algorithm to be used. Currently supported values are "RcppML" and "NMF".

tol

The tolerance for convergence (only applicable when nmf.method is "RcppML").

maxit

The maximum number of iterations for convergence (only applicable when nmf.method is "RcppML").

rev.nmf

Whether to perform reverse NMF (i.e., transpose the input matrix) before running the analysis.

ndims.print

The dimensions (number of basis vectors) to print in the output.

nfeatures.print

The number of features to print in the output.

reduction.name

Reduction to be stored in the Seurat object.

reduction.key

The prefix for the column names of the basis vectors.

verbose

Whether to print the message. Default is TRUE.

seed.use

Random seed.

cores

The number of threads to be used in older RcppML releases that expose a global OpenMP thread setter. Newer releases manage their thread count internally.

Examples

library(Matrix)
data(pancreas_sub)
pancreas_sub <- RunStandardWorkflow(pancreas_sub)
#> ℹ [2026-08-30 05:29:30] Start standard processing workflow...
#> ℹ [2026-08-30 05:29:30] Checking a list of <Seurat>...
#> ! [2026-08-30 05:29:30] Data 1/1 of the `srt_list` is "unknown"
#> Warning: Data 1/1 of the `srt_list` is "unknown"
#> ℹ [2026-08-30 05:29:30] Perform `NormalizeData()` with `normalization.method = 'LogNormalize'` on 1/1 of `srt_list`...
#> ℹ [2026-08-30 05:29:31] Perform `FindVariableFeatures()` on 1/1 of `srt_list`...
#> ℹ [2026-08-30 05:29:31] Use the separate HVF from `srt_list`
#> ℹ [2026-08-30 05:29:31] Number of available HVF: 2000
#> ℹ [2026-08-30 05:29:31] Finished check
#> ℹ [2026-08-30 05:29:31] Perform `ScaleData()`
#> ℹ [2026-08-30 05:29:31] Perform pca linear dimension reduction
#> ℹ [2026-08-30 05:29:32] Use stored estimated dimensions 1:23 for Standardpca
#> ℹ [2026-08-30 05:29:32] Perform `Seurat::FindClusters()` with `cluster_algorithm = 'louvain'` and `cluster_resolution = 0.6`
#> ℹ [2026-08-30 05:29:32] Reorder clusters...
#> ℹ [2026-08-30 05:29:33] Skip `log1p()` because `layer = data` is not "counts"
#> ℹ [2026-08-30 05:29:33] Perform umap nonlinear dimension reduction
#> ✔ [2026-08-30 05:29:42] Standard processing workflow completed
pancreas_sub <- RunNMF(pancreas_sub)
#> ℹ [2026-08-30 05:29:42] Running NMF...
#> ℹ BE_ 1 
#> ℹ Positive:  Ccnd1, Spp1, Eno1, Rps2, Mdk, Ldha, Pebp1, Mif, Gapdh, Prdx1 
#> ℹ 	   Cd24a, Krt8, Dbi, Aldoa, Cldn10, Rplp1, Mgst1, Npm1, Ybx1, Tkt 
#> ℹ 	   Hspe1, Ptma, Rpl12, Clu, Wfdc2, Sox9, Vim, Rpl22l1, Krt18, Tmsb10 
#> ℹ Negative:  Tmem108, Poc1a, Epn3, Wipi1, Tmcc3, Fgf12, Tecpr2, Zbtb4, Plekho1, Slc25a10 
#> ℹ 	   Gm10941, Trf, Man1c1, Hmgcs1, Nipal1, Jam3, Pgap1, Alpl, Tnr, Kcnip3 
#> ℹ 	   Gm15915, Rbp2, Cbfa2t2, Sh2d4a, Megf6, Prkcb, Naaladl2, Fam46d, Hist2h2ac, Tox2 
#> ℹ BE_ 2 
#> ℹ Positive:  Spp1, Atp1b1, Id2, Vim, Sparc, Dbi, Rps2, Mgst1, Gsta3, Rpl22l1 
#> ℹ 	   Acot1, Anxa2, 1700011H14Rik, Sox9, Rplp1, Bicc1, Adamts1, Rps12, Nudt19, Cldn3 
#> ℹ 	   Pdzk1ip1, Ppp1r1b, Rpl36a, Rpl12, Ifitm2, Jun, Ptn, S100a10, Mdk, Mt1 
#> ℹ Negative:  Aacs, Tmem108, Poc1a, Epn3, B830012L14Rik, Tmcc3, Rfc1, Wsb1, Tecpr2, Zbtb4 
#> ℹ 	   Plekho1, Ppp2r2b, Haus8, Trf, Gm5420, Man1c1, Hmgcs1, Nipal1, Jam3, Tcerg1 
#> ℹ 	   Pgap1, Alpl, Larp1b, Tnr, Kcnip3, Lsm12, Ptbp3, Gm15915, Cntln, Rbp2 
#> ℹ BE_ 3 
#> ℹ Positive:  Clu, Mt1, Spp1, Krt18, Mt2, Muc1, Gsta3, Cldn3, Mgst1, Sparc 
#> ℹ 	   Ttr, Serpinh1, Aldoa, Tpi1, H19, Dbi, Epcam, Mif, Pdzk1ip1, Gapdh 
#> ℹ 	   Prdx1, Cat, Ambp, Ldha, Ifitm2, Krt8, Csrp2, Anxa2, Rps2, Tmsb10 
#> ℹ Negative:  Rpa3, Aacs, Tmem108, Poc1a, Nop56, Wipi1, Tmcc3, Trib1, Rrp15, Fgfrl1 
#> ℹ 	   Nhsl1, Fgf12, Lama1, Slc20a1, Tecpr2, Zbtb4, Plekho1, Ppp2r2b, Haus8, Gm10941 
#> ℹ 	   Trf, Gm5420, Man1c1, Hmgcs1, Nipal1, Jam3, Tcerg1, Snrpa1, Alpl, Larp1b 
#> ℹ BE_ 4 
#> ℹ Positive:  Cck, Tmsb4x, Mdk, Gadd45a, Selm, Ppp1r14a, Sox4, Btbd17, Cotl1, Ppp3ca 
#> ℹ 	   Sh3bgrl3, Jun, Ifitm2, Rps2, Tubb3, Rplp1, Igfbpl1, Ptma, Calr, Sult2b1 
#> ℹ 	   Hn1, Lrpap1, Rps12, Smarcd2, Nkx6-1, Rpl36a, Neurog3, Neurod2, Gnas, Rpl12 
#> ℹ Negative:  Elovl6, Tmem108, Poc1a, Epn3, Nop56, Wipi1, B830012L14Rik, Atic, Rrp15, Rfc1 
#> ℹ 	   Fgf12, Lama1, Slc20a1, Tecpr2, Ppp2r2b, Eif1ax, Slc25a10, Fam162a, P4ha3, Gm10941 
#> ℹ 	   Tenm4, Pde4b, Gm5420, Man1c1, Hmgcs1, Pgap1, Mgst2, Larp1b, Tnr, Kcnip3 
#> ℹ BE_ 5 
#> ℹ Positive:  Spp1, Tmsb4x, Cyr61, Krt18, Tpm1, Krt8, Myl12a, Anxa5, Csrp1, Tnfrsf12a 
#> ℹ 	   Vim, Krt19, Jun, Anxa2, Tagln2, Nudt19, Sparc, Cldn7, Tmsb10, Clu 
#> ℹ 	   Myl12b, S100a10, Cd24a, Rps2, Myl9, Klf6, Cldn3, Lurap1l, Dbi, Rplp1 
#> ℹ Negative:  Elovl6, Aacs, Tmem108, Poc1a, Tmcc3, Rfc1, Nhsl1, Fgf12, Lama1, Slc20a1 
#> ℹ 	   Tecpr2, Plekho1, Ppp2r2b, Slc25a10, Fam162a, Gm10941, Tenm4, Man1c1, Nipal1, Jam3 
#> ℹ 	   Pgap1, Alpl, Tnr, Kcnip3, Ptbp3, Gm15915, Cntln, Ocln, Blvrb, Fras1 
#> ✔ [2026-08-30 05:29:59] NMF compute completed
CellDimPlot(
  pancreas_sub,
  group.by = "CellType",
  reduction = "nmf"
)


FeatureDimPlot(
  pancreas_sub,
  features = c("BE_1", "BE_2", "BE_3"),
  reduction = "UMAP",
  palette = "RdBu",
  xlab = "UMAP_1",
  ylab = "UMAP_2",
  theme_use = "theme_blank"
)


ht_cells <- NMFHeatmap(
  pancreas_sub,
  plot_type = "cells",
  cell_annotation = "CellType"
)
#> ℹ [2026-08-30 05:29:59] `NMFHeatmap()` input: 1000 cells x 50 NMF dimensions. Computing a 1000 x 1000 similarity matrix (~0.01 GiB dense numeric matrix).
#> ℹ [2026-08-30 05:29:59] Ordering `NMFHeatmap()` rows and columns ...
#> ℹ [2026-08-30 05:30:00] Building ComplexHeatmap object for `NMFHeatmap()` ...
#> ℹ [2026-08-30 05:30:00] Calculating `NMFHeatmap()` render size ...
#> ℹ [2026-08-30 05:30:00] Drawing `NMFHeatmap()`; this can take time for large similarity matrices ...
#> ℹ [2026-08-30 05:30:01] Assembling `NMFHeatmap()` plot object ...
ht_cells$plot


ht_features <- NMFHeatmap(
  pancreas_sub,
  plot_type = "features"
)
#> ℹ [2026-08-30 05:30:01] `NMFHeatmap()` input: 2000 features x 50 NMF dimensions. Computing a 2000 x 2000 similarity matrix (~0.03 GiB dense numeric matrix).
#> ℹ [2026-08-30 05:30:01] Ordering `NMFHeatmap()` rows and columns ...
#> ℹ [2026-08-30 05:30:01] Building ComplexHeatmap object for `NMFHeatmap()` ...
#> ℹ [2026-08-30 05:30:01] Calculating `NMFHeatmap()` render size ...
#> ℹ [2026-08-30 05:30:02] Drawing `NMFHeatmap()`; this can take time for large similarity matrices ...
#> ℹ [2026-08-30 05:30:16] Assembling `NMFHeatmap()` plot object ...
ht_features$plot