ct.normalizeFQ: Apply Factored Quantile Normalization to an eset

In RussBainer/gCrisprTools: Suite of Functions for Pooled Crispr Screen QC and Analysis

Apply Factored Quantile Normalization to an eset

Description

This function applies quantile normalization to subsets of samples defined by a provided factor, correcting for library size. It does this by converting raw count values to log2 counts per million and optionally adjusting further in the usual way by dividing these values by user-specified library size factors; then this matrix is split into groups according to the provided factor that are quantile normalized, and then the groups are median scaled to each other before conversion back into raw counts. This method is best used in comparisons for long timecourse screens, where groupwise differences in growth rate cause uneven intrinsic dialation of construct distributions.

Note that this normalization strategy is not appropriate for experiments where significant distortion of the libraries is expected as a consequence of the screening strategy (e.g., strong selection screens).

Usage

ct.normalizeFQ(eset, sets, lib.size = NULL)

Arguments

eset	An ExpressionSet containing, at minimum, count data accessible by exprs.
sets	A character or factor object delineating which samples should be grouped together during the normalization step. Must be the same length as the number of columns in the provided eset, and cannot contain 'NA' or 'NULL' values.
lib.size	An optional vector of voom-appropriate library size adjustment factors, usually calculated with calcNormFactors and transformed to reflect the appropriate library size. These adjustment factors are interpreted as the total library sizes for each sample, and if absent will be extrapolated from the columnwise count sums of the exprs slot of the eset.

Value

A renormalized ExpressionSet object of the same type as the provided object.

Author(s)

Russell Bainer

Examples

data('es')

#Build the sample key and library sizes for visualization
library(Biobase)
sk <- relevel(as.factor(pData(es)$TREATMENT_NAME), 'ControlReference')
names(sk) <- row.names(pData(es))
ls <- colSums(exprs(es))

es.norm <- ct.normalizeFQ(es, sets = gsub('(Death|Control)', '', pData(es)$TREATMENT_NAME), lib.size= ls)
ct.gRNARankByReplicate(es, sampleKey = sk, lib.size= ls)
ct.gRNARankByReplicate(es.norm, sampleKey = sk, lib.size= ls)

RussBainer/gCrisprTools documentation built on Nov. 5, 2022, 2:35 p.m.