Human Gene 1.0 ST array and exon arrays have only PM probesets and no MM probesets, thus generally it is not possible to perform MAS5 with these arrays. However, R package "xps" can indeed apply MAS5 to exon arrays and Human Gene 1.0 ST array. What is the detailed algorithm of MAS5 in "xps" package when it process these arrays? What are the differences between the modified MAS5 algorithm of "xps" and the regular MAS5 algorithm mentioned in sadd_whitepaper.pdf (Affymetrix website)?

Hi,

For IV arrays the MAS5 algorithm implemented in 'xps' is identical to the MAS5 
algorithm described in 'sadd_whitepaper.pdf'. Please have a look at Fig.4 of 
Chapter 2.2 of vignette 'APTvsXPS.pdf' where I compare the different implementations 
of MAS 5, i.e. Expression Console (EC), APT, xps, and affy. 

To understand why 'xps' is able to implement MAS5 for whole genome and exon arrays
you need to know how Affymetrix does calculate the background intensities for the
different arrays: 

- For IV arrays Affymetrix does not simply subtract MM from PM for 
each probe, but first calculates an 'Ideal Mismatch' (IM), which is then subtracted 
from PM.

- For whole genome and exon arrays Affymetrix does not use MM probes but uses 
special ”genomic” and ”antigenomic” probes to compute an IM.

Affymetrix explains this already in 'Expression analysis algorithm tutorial.pdf'
for the original MAS4 algorithm, see Chapter III.c. Background Calculation. As
you see the array is divided into 4x4 sectors, and the IM is calculated for each
sector. 
Please read Chapter 3.2 (sector) of my vignette 'xpsPreprocess.pdf'.

In their 'sadd_whitepaper.pdf' for MAS5 Affymetrix explains that to calculate the
Ideal Mismatch (IM) they use the one-step biweight algorithm (page 6). Once again 
the array is divided into 4x4 sectors (see the Figure on page 4). 
Please read Chapter 3.3 (weightedsector) of my vignette 'xpsPreprocess.pdf'.

For whole genome and exon arrays Affymetrix has created background probes with 
different GC-contents, see their 'exon_background_correction_whitepaper.pdf'.
However, these probes are now used to compute a background for PM probes with 
the same GC-content. 
Please read Chapter 3.4 (gccontent) of my vignette 'xpsPreprocess.pdf'.

In summary, this means that I can compute the background intensities separately
for each type of array and then use the calculated background probes for further
processing. In the case of the whole genome and exon arrays I use simply the
background intensities computed with the GC-content.

Best regards,
Christian