M3D_Get_Extremes: Get outliers from MM curve.
In tallulandrews/M3D: Michaelis-Menten Modelling of Dropouts in single-cell RNASeq

Description Usage Arguments Details Value Examples

Identifies outliers left and right of a fitted Michaelis-Menten curve. Functions tagged with "bg__" are not meant for direct usage and are not available in the Bioconductor release.

1
2

  M3DropGetExtremes(expr_mat, fdr_threshold=0.1, percent=NA, v_threshold=c(0.05,0.95), suppress.plot=FALSE)
  bg__get_extreme_residuals(expr_mat, fit=NA, fdr_threshold=0.1, percent=NA, v_threshold=c(0.05, 0.95), direction="right", suppress.plot=FALSE)

`expr_mat`	a numeric matrix of normalized (not log-transformed) expression values, columns = samples, rows = genes.
`fit`	output from fitting the Michaelis-Menten equation, see `bg__fit_MM`
`fdr_threshold`	the threshold for identifying significant outliers after multiple testing correction.
`percent`	identify this percentage of data that is most extreme in each direction.
`v_threshold`	restrict to this range of dropout rates to avoid poorly fit regions of the data.
`direction`	"left" or "right", which tail of outliers to examine.
`suppress.plot`	logical, whether to plot the fitted Michaelis-Menten curve and highlight to identified most extreme outliers.

Fits a Michaelis-Menten function to the dropout-rate of the provided data, then identifies the most extreme left and/or right outliers from the curve. Horizontal residuals are calculated as :

log_10(S) - log_10( (K * (1 - P)) / P )

. Extreme left[right] outliers are identified either as the percent smallest[largest] horizontal residuals. If percent is undefined (default) a normal distribution is fitted to the horizontal residuals and a Z-test is used to identify significant outliers after FDR multiple testing correction.

Only genes with dropout rates within v_threshold will be considered to avoid the skewing of residuals due to the exponential parts of the MM curve near P = 0 & P = 1.

M3DropGetExtremes identifies both left and right residuals using the provided thresholds in each direction. Eg. will return the percent smallest and percent largest residuals. It also plots the fitted MM curve and highlights the left and right extreme outliers unless suppress.plot=TRUE . bg__get_extreme_residuals identifies one-sided extreme outliers.

M3DropGetExtremes List containing elements left and right, vectors of the names of the extreme genes to the left and right of the curve respectively. bg__get_extreme_residuals A vector of names of the extreme genes in the specified direction.

  library(M3DExampleData)
  norm <- M3DropConvertData(Mmus_example_list$data, is.counts=TRUE)
  extreme_gene_lists <- M3DropGetExtremes(norm, fdr_threshold=0.1)
  extreme_gene_lists <- M3DropGetExtremes(norm, percent=0.01)
#  Lextremes <- bg__get_extreme_residuals(norm, fdr_threshold=0.1, direction="left")