MiNorm {MiDA}

Microarray data normalization

Description

Normalizes microarray expression intensities using different methods with or without background correction.

Usage

MiNorm(Matrix, posNC, method = "none", leaveNC = TRUE, BGcor = FALSE)

Arguments

Matrix

numeric matrix of intensities data where each row corresponds to a probe (gene, transcript), and each column correspondes to a specimen (patient).

posNC

numeric vector specifying numbers of rows containing negative controls (non-coding areas). Used for method="SQN" only. Rows with negative controls will be removed from an intensity matrix after the normalization if leaveNC=FALSE.

method

character string specifying normalization method. Possible values are: "none" (no normalization) "center" (subtracting the row mean), "scale" (dividing by row standard deviation), "standardize" (subtracting the row mean and dividing by row standard deviation - z-score transformation), "range" (ranges from 0 to 1), "QN" (normalization based upon quantiles), "SQN" (subset quantile normalization using negative control features), "Loess" (cyclicly applying loess normalization).

leaveNC

logical value indicating whether rows with negative control should be deleted from intensity matrix after normalization.

BGcor

logical value indicating whether background correction should be done before normalization. Could be used for background correction only (without data normalization) if method="none".

Details

This function is intended to normalize microarray intensities data between arrays. Background correction is optional.
Background correction method is "normexp", which is based on a convolution model (Ritchie, 2007). See backgroundCorrect for details.
Quantile normalization method implies that we can give each array the same distribution See normalize.quantiles for details.
Subset quantile normalization is performed based on a subset of negative (or non-coding) controls according to (Wu and Aryee, 2010). Number of normal distributions in the mixture approximation is 5, weight given to the parametric normal mixture model is 0.9. See SQN for details.
Cyclic loess normalization implements method of Ballman et al (2004), whereby each array is normalized to the average of all the arrays. See normalizeCyclicLoess for details.

Value

A matrix of the same dimensions as Matrix containing normalized values with or without background correction. If leaveNC=FALSE the function returns a matrix with normalized values without rows containing negative controls.

Author(s)

Elena N. Filatova

References

Ballman K.V., Grill D.E., Oberg A.L. and Therneau T.M. (2004). Faster cyclic loess: normalizing RNA arrays via linear models. Bioinformatics 20, 2778-2786. https://doi.org/10.1093/bioinformatics/bth327

Bolstad B.M., Irizarry R.A., Astrand M. and Speed T.P. (2003) A Comparison of Normalization Methods for High Density Oligonucleotide Array Data Based on Bias and Variance. Bioinformatics 19(2), 185-193. https://doi.org/10.1093/bioinformatics/19.2.185

Ritchie M.E., Silver J., Oshlack A., Silver J., Holmes M., Diyagama D., Holloway A. and Smyth G.K. (2007). A comparison of background correction methods for two-colour microarrays. Bioinformatics 23, 2700-2707. https://doi.org/10.1093/bioinformatics/btm412

Wu Z and Aryee M. (2010). Subset Quantile Normalization using Negative Control Features. Journal of Computational Biology 17(10), 1385-1395. https://doi.org/10.1089/cmb.2010.0049

See Also

backgroundCorrect, normalizeCyclicLoess, normalize.quantiles, SQN

Examples

data("IMexpression")
# Loess normalization
LoMatrix<-MiNorm(IMexpression, method="Loess")
par(mfrow=c(1,2))
boxplot(log2(IMexpression),main="Before normalization")
boxplot(log2(LoMatrix),main="Loess normalization")
par(mfrow=c(1,1))