A Bayesian calibration model for combining different pre-processing methods in Affymetrix chips

Table 4 Annotation for differentially expressed genes

First 292 genes with the largest posterior probability
	Combined Model	MAS5	RMA	dChip
Biological Processes	181	146	171	168
Molecular Functions	193	154	180	173
Cellular Components	179	145	173	169
KEGG pathways	223	182	215	205

The table presents the number of differentially expressed genes annotated by GO or KEGG considering the first 292 probesets with the largest tail posterior probability for each method and for the combined one. The most represented KEGG pathways are mainly related to immune response and oxidation (Antigen processing and presentation, MAPK signalling pathway, PPAR signaling pathway). For these pathways the number of genes found by the combined method is always larger than the one found by each pre-processing method. Moreover the combined model also calls 6 genes that are involved in the Insulin signalling pathway, specifically related to diabetes, one of the diseases recently highlighted to be associated with high fat diet in mice (see for example [45] and [46]). The same pathway is present with only 3 genes in MAS5 and with 2 genes in RMA and dChip.