3. Conclusions

Abstract

D. Diez et al. 2 Fig. 1. A) MA plot comparing two samples with different probe types highlighted. B) MA plot for the same samples but using the SNR information to label probes. If the SNR is greater than one in the two samples, the spots are black, orange if one of the SNR is less than one and red when both are. C) Array pseudo image showing the background intensities indicated as increasing levels of blue. In particular, the default method available on the Codelink software, median normalization, has the disadvantage that it does not take this behavior into account. For this reason, several normalization methods have been included in the codelink package allowing easy comparisons. In addition to median normalization, quantile, and CyclicLoess are available. In several studies (Shi, et al., 2006; Wu, et al., 2005) comparing nor-malization methods in the Codelink platform, better performance has been found using quantile and CyclicLoess compared to median normalization. On the other hand, CyclicLoess performed slightly better compared to quantile normalization (Wu, et al., 2005). Several diagnostic plots are available, exploiting the singular characteristics of the Codelink platform. Density plots can be useful for assessing pre-processing steps, such as background correction and normalization. MA plots have been widely used to assess normalization performance in two channel microarrays, but are also very valuable in one channel microarrays. It is possible to highlight spots based on the spot type (e.g. discovery probes, positive probes, negative probes, etc.) or SNR (Figure 1A and 1B). Other useful plots are the array pseudo images representing probe intensities , background or SNR into the chip location (Figure 1C). These can be used to detect spot artifacts and to assess the effect of background correction and normalization. In addition to the codelink package, annotation packages for the different arrays are provided. These packages, built using AnnBuilder (Zhang, 2006), contain information about the array probes with links to several biological databases like GeneOntology, KEGG, Uniprot, etc. The annotation packages are updated every Bioconductor release, allowing an up-to-date description of the gene products corresponding to each probe. In addition , there are utility functions to help write report files from the list of differentially expressed genes, including the corresponding information from the annotation packages. As with every package in the Bioconductor project, the codelink package comes with documentation in form of a vignette file. For further details about the package …