Normalizing From the Same Platform

If the arrays of interest are of the same platform, follow the same procedure as outlined in normalizing microarrays.

Normalizing From Different Platforms

If the samples for normalizing are from different platforms,

Quantile

1. Normalize arrays that belong to the same platform together
2. Once normalized platform-specific, quantile normalize to produce the same distribution across all arrays.
   • Quantile is the easiest to perform and will ensure the samples have the same distribution. Only genes common to all platforms should be selected. After normalization, check if normalization worked by clustering the samples, or by PCA. If samples cluster by a batch effect (eg. source) or platform, then quantile normalization was not adequate. For discussion on cross-platform normalization see,
     • ​Three methods for optimization of cross-laboratory and cross-platform microarray expression data
     • ​Genome Technology - How do you compare array data across different platforms?

  library(limma)
  # Read the data
  data = read.delim(myFile)
  # Quantile normalize 
  data.quantile_normalize = normalizeQuantiles(data[,c(2:ncol(data))])

Other

• The above method may not remove batch effects (ie. samples clustering by platform)
• Several methods are available to remove batch effects:
  • ​Removing batch effects in analysis of expression microarray data: an evaluation of six batch adjustment methods.
  • ​Adjusting batch effects in microarray expression data using empirical Bayes methods, as implemented by ​ComBat, is recommended in the study comparing several methods for removing batch effects.
  • ​virtualArray: a R/bioconductor package to merge raw data from different microarray platforms