Skip to contents

[Experimental] This is an experimental function and is subject to change.

Performs consensus integrative non-negative matrix factorization (c-iNMF) to return factorized \(H\), \(W\), and \(V\) matrices. In order to address the non-convex nature of NMF, we built on the cNMF method proposed by D. Kotliar, 2019. We run the regular iNMF multiple times with different random starts, and cluster the pool of all the factors in \(W\) and \(V\)s and take the consensus of the clusters of the largest population. The cell factor loading \(H\) matrices are eventually solved with the consensus \(W\) and \(V\) matrices.

Please see runINMF for detailed introduction to the regular iNMF algorithm which is run multiple times in this function.

The consensus iNMF algorithm is developed basing on the consensus NMF (cNMF) method (D. Kotliar et al., 2019).

Usage

runCINMF(object, k = 20, lambda = 5, rho = 0.3, ...)

# S3 method for liger
runCINMF(
  object,
  k = 20,
  lambda = 5,
  rho = 0.3,
  nIteration = 30,
  nRandomStarts = 10,
  HInit = NULL,
  WInit = NULL,
  VInit = NULL,
  seed = 1,
  nCores = 2L,
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

# S3 method for Seurat
runCINMF(
  object,
  k = 20,
  lambda = 5,
  rho = 0.3,
  datasetVar = "orig.ident",
  layer = "ligerScaleData",
  assay = NULL,
  reduction = "cinmf",
  nIteration = 30,
  nRandomStarts = 10,
  HInit = NULL,
  WInit = NULL,
  VInit = NULL,
  seed = 1,
  nCores = 2L,
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

Arguments

object

A liger object or a Seurat object with non-negative scaled data of variable features (Done with scaleNotCenter).

k

Inner dimension of factorization (number of factors). Generally, a higher k will be needed for datasets with more sub-structure. Default 20.

lambda

Regularization parameter. Larger values penalize dataset-specific effects more strongly (i.e. alignment should increase as lambda increases). Default 5.

rho

Numeric number between 0 and 1. Fraction for determining the number of nearest neighbors to look at for consensus (by rho * nRandomStarts). Default 0.3.

...

Arguments passed to methods.

nIteration

Total number of block coordinate descent iterations to perform. Default 30.

nRandomStarts

Number of replicate runs for creating the pool of factorization results. Default 10.

HInit

Initial values to use for \(H\) matrices. A list object where each element is the initial \(H\) matrix of each dataset. Default NULL.

WInit

Initial values to use for \(W\) matrix. A matrix object. Default NULL.

VInit

Initial values to use for \(V\) matrices. A list object where each element is the initial \(V\) matrix of each dataset. Default NULL.

seed

Random seed to allow reproducible results. Default 1.

nCores

The number of parallel tasks to speed up the computation. Default 2L. Only supported for platform with OpenMP support.

verbose

Logical. Whether to show information of the progress. Default getOption("ligerVerbose") or TRUE if users have not set.

datasetVar

Metadata variable name that stores the dataset source annotation. Default "orig.ident".

layer

For Seurat>=4.9.9, the name of layer to retrieve input non-negative scaled data. Default "ligerScaleData". For older Seurat, always retrieve from scale.data slot.

assay

Name of assay to use. Default NULL uses current active assay.

reduction

Name of the reduction to store result. Also used as the feature key. Default "cinmf".

Value

  • liger method - Returns updated input liger object

    • A list of all \(H\) matrices can be accessed with getMatrix(object, "H")

    • A list of all \(V\) matrices can be accessed with getMatrix(object, "V")

    • The \(W\) matrix can be accessed with getMatrix(object, "W")

  • Seurat method - Returns updated input Seurat object

    • \(H\) matrices for all datasets will be concatenated and transposed (all cells by k), and form a DimReduc object in the reductions slot named by argument reduction.

    • \(W\) matrix will be presented as feature.loadings in the same DimReduc object.

    • \(V\) matrices, an objective error value and the dataset variable used for the factorization is currently stored in misc slot of the same DimReduc object.

References

Joshua D. Welch and et al., Single-Cell Multi-omic Integration Compares and Contrasts Features of Brain Cell Identity, Cell, 2019

Dylan Kotliar and et al., Identifying gene expression programs of cell-type identity and cellular activity with single-cell RNA-Seq, eLife, 2019

Examples

# \donttest{
pbmc <- normalize(pbmc)
#>  Normalizing datasets "ctrl"
#>  Normalizing datasets "stim"
#>  Normalizing datasets "stim" ... done
#> 
#>  Normalizing datasets "ctrl"

#>  Normalizing datasets "ctrl" ... done
#> 
pbmc <- selectGenes(pbmc)
#>  Selecting variable features for dataset "ctrl"
#>  ... 168 features selected out of 249 shared features.
#>  Selecting variable features for dataset "stim"
#>  ... 166 features selected out of 249 shared features.
#>  Finally 173 shared variable features are selected.
pbmc <- scaleNotCenter(pbmc)
#>  Scaling dataset "ctrl"
#>  Scaling dataset "ctrl" ... done
#> 
#>  Scaling dataset "stim"
#>  Scaling dataset "stim" ... done
#> 
if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    pbmc <- runCINMF(pbmc)
}
#> Replicating iNMF runs ■■■■■■■■■■■■■■■■                  50% | [ 5 / 10 ] ETA:  …
#> Replicating iNMF runs ■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■  100% | [ 10 / 10 ] ETA: …
#>  Taking the consensus
#>  Taking the consensus ... done
#> 
#>  ANLS optimization with consensus fixed
#>  ANLS optimization with consensus fixed  ... objective error: 35854.9654727417
#> 
# }