[BioC] Does the strand of a microarray probe matter?

[Maria Stalteri]

Hi Cei,

I think you may still be a little confused about what the gene expression 
arrays are designed to measure.

The probes on the Affymetrix 3' expression arrays are single-stranded 
oligos having the sense sequence, i.e. the same sequence as the mRNA they are designed
 to detect, and the IVT assay used with these arrays produces single 
stranded cRNAs with the antisense sequence, i.e. the reverse complement
of the initial mRNA sample.

The probes on the Affymetrix Exon ST and Gene ST  arrays have the 
antisense sequence, i.e. the reverse complement of the sequence they are 
designed to detect, and the WT assay used with these arrays produces a 
single-stranded cDNA with the sense sequence, i.e. the same sequence
as the initial mRNA sample.

The probesets on these arrays are only designed to measure 
expression from one strand.

They will only measure expression from both strands in the cases where 
Affymetrix have tiled probesets on both strands in the same region
of the genome. This is the case for some probesets that were designed
based on ESTS, where it wasn't clear which strand the gene was on at the
time of array design, so probesets were tiled on both strands in the 
region the EST mapped to.


As for problems with probeset annotations or discrepancies between one 
annotation source and another, we have also found that
the number of annotation errors is probably somewhere close to 10%, 
and that genes that were close together or had overlapping ends tended
to cause problems for the annotations.

Affymetrix grades the reliability of the annotations for the 3' expression 
arrays as A, B, C or E for each probeset, with A being the most reliable 
and E being annotations based on EST clusters and generally the least 
reliable. We have found that although their A and B grade annotations are 
not always correct either, they are indeed more likely to be correct than 
the annotations they label as E.

For the exon arrays, Affymetrix labels its probesets as unique, similar, 
or mixed depending on whether or to what extent the probes 
cross-hybridise, so that one can choose to use only those probesets 
labelled as unique if one wants to avoid cross-hybridising probes.
(I haven't done any mappings of the probes on the exon arrays yet, so I 
don't know how true this is.)

Best wishes,
Maria