Statistics and Data Analysis for Microarrays Using R and Bioconductor, 2nd Edition

Data preprocessing and n ormalization 721

x

′

ki

= 2

A

k

+

M

′

k

2

(20.37)

x

′

kj

= 2

A

k

−

M

′

k

2

(20.38)

The preferr e d method is to calculate the normalization curves using rank

invariant sets of probes [59].

Since a typical Affymetrix e xperiment involves more than two arrays, the

method applies the above normalization for every pair of two arrays from the

set. If there are n arrays in the data set, every array k will re quire n − 1

adjustments: 1, 2, . . . , k − 1, k + 1, . . . n. These adjustments are all weighted

equally and applied to the set of arrays. This entire process is r epeated until

the calculated adjustments are below a certain threshold, at which point ...