tuneFilterWeights: Tune filter weights

In PFPrzytycki/CellWalkR: An R Package for Integrating and Visualizing Single-Cell and Bulk Data to Resolve Regulatory Elements

Tune filter weights

Description

tuneFilterWeights generates cell homogeneity scores for filter weight settings (by default 0 and 1 for on and off) across all filters. Each filter can either apply just to specific peaks or to whole genes (filterGene, TRUE by default) and can be permissive or filtering out (filterOut, FALSE by default). Regions mapping peaks to genes must be provided to use filters. If filters are applied to specific peaks, peaks for the ATAC data must also be provided. To speed up calculations cells can be subsampled to a specified sampleDepth (default 5000).

Usage

tuneFilterWeights(
  cellEdges,
  labelGenesList,
  labelEdgesList,
  labelEdgeWeights,
  ATACMat,
  ATACGenePeak,
  method = "Expression",
  filters,
  filterWeightOpts = c(0, 1),
  filterOut,
  filterGene,
  regions,
  peaks,
  sampleDepth = 5000,
  steps,
  cellTypes,
  parallel = FALSE,
  numCores = 1,
  trackProgress = FALSE,
  tensorflow = FALSE
)

Arguments

cellEdges	matrix of cell-to-cell similarity (c-by-c)
labelGenesList	list of tables with genes of interest in first column and corresponding labels in second column, optionally log-fold change in expression values in the third
labelEdgesList	list of cell-to-label similarity matrices each with dimensions c-by-l
labelEdgeWeights	vector of edge weights corresponding to list of cell-to-label similarity matrices
ATACMat	either a cell-by-peak matrix or a SnapATAC object
ATACGenePeak	per label mapping of peaks to genes returned by mapPeaksToGenes()
method	scaling method, either "Expression","Correlation", or "None"
filters	list of GRanges for locations that pass filters
filterWeightOpts	vector of parameters to test for weights assigned to each filter
filterOut	boolean for each filter of whether those regions are permitted or filtered out
filterGene	boolean for each filter of whether it applies to genes as a whole or just overlapping peaks (note that weights do not apply to peaks)
regions	regions map peaks to genes
peaks	GRanges of peaks in ATACMat
sampleDepth	integer, depth to subsample cells to for faster calculation
steps	integer indicating number of steps to take if walk should not be run to convergence
cellTypes	list of names of cell types, if not provided unique cell labels are used
parallel	boolean, execute in parallel
numCores	integer, number of cores to use for parallel execution
trackProgress	boolean, print percent run completed
tensorflow	boolean to indicate whether to compute on GPU

Value

data frame of parameters with corresponding cell homogeneity

Examples

data("SampleCellWalkRData")
regions <- getRegions()
filters <- list(SampleCellWalkRData$filter)
labelGenesList <- list(SampleCellWalkRData$labelGenes)
labelEdgesList <- list(SampleCellWalkRData$labelEdges)
## Not run: tuneFilterWeights(SampleCellWalkRData$cellEdges,
                  labelGenesList,
                  labelEdgesList,
                  1,
                  SampleCellWalkRData$ATACMat,
                  SampleCellWalkRData$ATACGenePeak,
                  filters=filters,
                  regions=regions)
## End(Not run)

PFPrzytycki/CellWalkR documentation built on Oct. 26, 2023, 1:50 p.m.