High-density feature of in situ synthesized microarray makes it possible to design multiple probes within a target gene to interrogate its transcriptional activity, and Affymetrix GeneChip is the precedent that uses a set of PM and MM probe pairs to represent each gene. Various methods of summarizing a gene’s expression intensity from its probe set have been proposed (Rafael (2002), Li and Wong (2001a), Naef et al. (2001), and Holder et al. (2001)). RMA (Robust Multi-Array Average) algorithm is the unique one that only relies upon the PM intensities and utilizes the information from replicate arrays. We used the same method with RMA algortihm to calculate a gene expression intensiy, and below shows how this procedure works in our testing dataset.

MedianPolish :: Gene expression intensity measuring module

To summarize the expression intensity of each target gene from the 15 data points, we perform Tukey’s median polish procedure to calculate a grand median intensity from the matrix of 5x3 probes’ intensities, which is strongly recommended by RMA. This procedure is based on additive linear model, as the following formula indicates that the observed intensity is assumed to be the summation of real hybridization intensity, plus probe affinity effect and experimental errors. By Tukey’s median polishing, after a series of iteration of residue subtraction, the matrix will come to a stable status, and the final number will be estimated as the M value. The following table shows how we estimate a grand median from a 5 x 3 data matrix.