Median-balanced quantile normalization

Description Usage Arguments Details Value Author(s) References See Also Examples

Modified quantile-normalization (QN) of a matrix, e.g., intensity values from omics data or other data sorted in columns. The modification prevents systematic flattening of features (rows) which are rank invariant (RI) or nearly rank invariant (NRI) across columns, for example features that populate mainly the tails of the intensity distribution or features that separate in intensity.

mbqn(
  x,
  FUN = "mean",
  na.rm = TRUE,
  method = "limma",
  offsetmatrix = FALSE,
  verbose = FALSE
)

`x`	a data matrix, where rows represent features, e.g. of protein abundance, and columns represent groups or samples, e.g. replicates, treatments, or conditions.
`FUN`	a function like mean, median (default), a user defined function, or a numeric vector of weights with length `nrow(x)` to balance each feature across samples. Functions can be parsed also as characters. If FUN = NULL, features are not balanced, i.e. normal QN is used.
`na.rm`	logical indicating to omit NAs in the computation of feature mean.
`method`	character specifying function for computation of quantile normalization; "limma" (default) for `normalizeQuantiles()` from the limma package or "preprocessCore" for `normalize.quantiles()` from the preprocessCore package.
`offsetmatrix`	logical indicating if offset matrix should be used instead of offset vector specifying offset for each row
`verbose`	logical indicating to print messages.

Balance each matrix row by substracting its feature offset computed with FUN, e.g. the median; apply quantile-normalization and add the feature means to the normalized matrix. For further details see [4]. For quantile normalization with the "limma" package see [1,2] and for the preProcessCore package see [3].

Normalized matrix

Ariane Schad

[1] Smyth, G. K., and Speed, T. P. (2003). Normalization of cDNA microarray data. Methods 31, 265<e2><80><93>273.
Ritchie, M.E., Phipson, B., Wu, D., Hu, Y., Law, C.W., Shi, W., and Smyth, [2] G.K. (2015). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Research 43(7), e47.
[3] Bolstad B. M. (2016). preprocessCore: A collection of pre-processing functions. R package version 1.36.0. https://github.com/bmbolstad/preprocessCore
[4] Brombacher, E., Schad, A., Kreutz, C. (2020). Tail-Robust Quantile Normalization. BioRxiv.

mbqnNRI(), mbqnGetNRIfeatures().

## Compute mean and median balanced quantile normalization
X <- matrix(c(5,2,3,NA,4,1,4,2,3,4,6,NA,1,3,1),ncol=3)
mbqn(X, mean) # Use arithmetic mean to center features
mbqn(X, median) # Use median to center features
mbqn(X, "median")

## Use user defined array of weights for averaging
wt <- c(1,3,1)/5 # Weights for each sample
user_array <- apply(X,1,weighted.mean, wt ,na.rm =TRUE)
mbqn(X, user_array)