M3D_Clean_Data: Filter Expression Data
In tallulandrews/M3D: Michaelis-Menten Modelling of Dropouts in single-cell RNASeq

Description Usage Arguments Details Value Examples

Filters and normalizes a given expression matrix. Removes low quality cells and undetected genes, and normalizes counts to counts per million. Functions tagged with "bg__" are not meant for direct usage and are not available in the Bioconductor release.

1
2

  M3DropCleanData(expr_mat, labels = NA, is.counts = TRUE, suppress.plot = FALSE, pseudo_genes = NA, min_detected_genes = NA)
  bg__filter_cells(expr_mat, labels = NA, suppress.plot = FALSE, min_detected_genes = NA)

`expr_mat`	a numeric matrix of raw or normalized (not log-transformed) expression values, columns = samples/cells, rows = genes.
`labels`	a vector of length equal to the number of columns of expr_mat with names or group IDs for each cell.
`is.counts`	logical, whether the provided data is unnormalized read/fragment counts.
`suppress.plot`	logical, whether to plot the distribution of number of detected genes per cell.
`pseudo_genes`	a vector of gene names of known pseudogenes which will be removed from the cleaned data.
`min_detected_genes`	minimum number of genes/cell for a cell to be included in the cleaned data.

Retains genes detected (expression>0) in more than 3 cells and with mean normalized expression >= 10^-5. If min_detected_genes is defined all cells not reaching the threshold are removed. Otherwise, fits a normal distribution to the distribution of detected genes/cell and removes those cells with significantly few detected genes (FDR 5%). This fit is plotted for visual inspection. If is.counts==TRUE then each column is converted to counts per million (ignoring ERCC spike-ins if present).

A list with elements: data, the normalized filtered expression matrix; and labels, labels of the remaining cells.

  library(M3DExampleData)
  # Remove all cells with < 2000 detected genes and convert to cpm 
  cpm <- M3DropCleanData(Mmus_example_list$data, Mmus_example_list$labels, 
			is.counts=TRUE, min_detected_genes=2000)
  # Removes cells with significantly few detected genes (FDR=5%)
  filtered_only <- M3DropCleanData(Mmus_example_list$data, Mmus_example_list$labels, 
			is.counts=FALSE)
#  QCed <- bg__filter_cells(Mmus_example_list$data[,1:10], Mmus_example_list$labels[1:10], suppress.plot=TRUE)