RMA and GC-RMA Normalisation
A comparison of the Robust Multiarray Averaging (RMA) method of microarray normalization and summarization with standard methods such as those implemented in the Affymetrix Microarray Suite (MAS). Affymetrix GeneChips are made up of a number of probes, each designed to measure the expression levels of a particular genomic sequence. Each probe consists of hundreds of short 25-mer oligonucleotide strands that match the target mRNA sequnce exactly. RNA samples are transformed into cRNA (complimentary RNA) using an in vitro transcription method that. cRNA is fragmented, biotin labels are attached, and the fragments are washed over the chip. cRNA fragments hybridise with their respective oliginucleotide sequence, resulting in an increase in abundance of biotinylated cRNA molecules on the probe. A fluorescent dye is washed over the chip and binds to the biotin labels. A fluorescent scan of the chip gives fluorescence values for each probe, which are directly related to the abundance of cRNA fragments hybridised to the probe. These fluorescence levels can be used to infer the relative abundance of specific mRNA sequences in a sample, giving a "snapshot" of the transcriptome.
Affymetrix Genechips are designed in such a way that each gene is matched to 11-20 such probes evenly distributed throughout the chip. These probes make up a Probe Set on the chip. The presence of noise is a large problem in large scale microarray studies, and can come from a number of sources. Noise due to non-specific binding of cRNA fragments to probes is one such source, and is eleviated through the use of probes on the chip designed to measure for non-specific binding. Each probe described above, termed a Perfect Match probe (PM) is matched to a second probe, termed a Mis-Match probe (MM). MM probes are identical to PM probes but for the middle (13th) nucleotide in the sequence, which is replaced with it's complement. By subtracting the signal for the MM probe from the signal for the PM probe, a value for the true signal is reached.
However, it has been shown that the signal strength for the MM probes can often be larger than that of the PM probes, implying that the MM probe is detecting true signal as well as background signal. This can result in non-sensical negative expression values. RMA is a normalisation procedure for microarrays that background corrects, normalises and summarises the probe level information without the use of the information obtained in the MM probes. GeneChip RMA (GC-RMA) is an improved form of RMA that is able to use the sequence-specific probe affinities of the GeneChip probes to attain more accurate gene expression values.
In this talk I discuss the merits and possible folies of using RMA normalisation compared to standard methods.
The presentation can be found here.