Imputation: Normalized Data using the DANB model

In tallulandrews/M3D: Michaelis-Menten Modelling of Dropouts in single-cell RNASeq

Description

Normalizes data to a common library size, imputing zeros as needed.

Usage

	NBumiImputeNorm(counts, fit, total_counts_per_cell=median(fit$vals$tis))

Arguments

counts	raw count matrix, rows=genes, cols=cells
fit	Output from NBumiFitModel
total_counts_per_cell	reference library size to normalize all cells to.

Details

Converts raw counts into positions in the CDF for the depth-adjusted negative binomial model fit to each observation. Adjusts the DANB parameters (mean and size) for the new library size. Then calculate the normalized counts for the equivalent position in the CDF for the NB using the new parameters.

Value

Normalized count matrix.

Examples

library(M3DExampleData)
counts <- as.matrix(Mmus_example_list$data);
counts <- counts[rowSums(counts) > 0,];
fit <- NBumiFitModel(counts);
normed_counts <- NBumiImputeNorm(counts, fit, 1000000)

tallulandrews/M3D documentation built on May 31, 2019, 2:55 a.m.