A significant number of probes that appeared to be partially or completely non-responsive are now known to exist, but methods of mitigating them are not adequate. This technique addresses these concerns about the effect of the individual oligonucleotide probes.
Known as REDI (REDuction of Invariant Probes) analysis, this method essentially removes or masks the affect caused by potentially invariant probes within a probe set in order to provide a more accurate indication of potential differential expression.
Affymetrix probe sets contain multiple oligonucleotide probes for interrogating transcripts. Through scientific research and an examination of thousands of hybridisation experiments, it has established that certain Perfect Match probes fail to respond adequately to the amount of target transcript in a sample.
Removal of these invariant probes from the calculations used to determine hybridisation signal intensity yields a more accurate determination of relative transcript abundance and increases the sensitivity of the probe set to variations in transcript levels.
By applying these algorithms to existing .cel files, increases in the rates of detection of differentially expressed transcripts without increases in the false discovery (positive) rates have been seen.
"We use CEL files with the chosen signal measure algorithm to augment differential gene lists for Human, Rat, and Mouse chips." said Steve Casey, Expression Analysis Founder and COO.
REDI can be used with any of the current signal measures such as RMA, PDNN, dChip, MAS5, or PLIER.
Expression Analysis recently conducted a review of HG-U133-based PM probes using thousands of hybridisations of GeneChips combined with RefSeq-based probe sequence analysis.
The study suggested that approximately 30 per cent of the HG-U133A PM probes were non-responsive or relatively invariant, affecting more than 80 per cent of the probe sets. Similar results were subsequently found for Rat and Mouse chips.
In addition, Expression Analysis found that the impact of these probes served to underestimate fold change differences or mask differential expression entirely when probesets contained a large number of these probes.
The REDI analysis, however, typically augments a differential gene list by 25 per cent - 200 per cent when the lists are based on significance and fold change thresholds.