Binned-Intensity Normalization Algorithm for Single-Dye Microarrays

Gene Cutler1, Tularik Inc

Changes in mRNA levels measured in microarray experiments are usually reported as ratios, where the mRNA levels in an experimental sample are divided by the mRNA levels in a reference sample. In order for this value to be meaningful, global differences between the data sets to be compared must be minimized through normalization. This is especially true when comparing data across multiple arrays, as is done in single-dye microarray systems, where array differences and hybridization differences come into play. While median normalization performs adequately when the differences between data sets are linear and the data is well behaved, it fails to produce good results when there are non-linearities in the data sets. To cope with this situation, I have implemented a binned-intensity normalization algorithm which corrects for non-linearities and noisy or variable data points, generating more reliable data than simple median normalization.