Microarray data are generated in complex experiments and frequently compromised by a variety of systematic errors. Subsequent data normalization aims to correct these errors. Although several normalization methods have recently been proposed, they frequently fail to account for the variability of systematic errors within and between microarray experiments. However, optimal adjustment of normalization procedures to the underlying data structure is crucial for the efficiency of normalization. To overcome this restriction of current methods, we have developed two normalization schemes based on iterative local regression combined with model selection. The schemes have been demonstrated to improve considerably the quality of normalization. They are implemented in a freely available R package. Additionally, functions for visualization and detection of systematic errors in microarray data have been incorporated in the software package. A graphical user interface is also available.
Availability: The R package can be downloaded from http://itb.biologie.hu-berlin.de/~futschik/software/R/OLIN. It underlies the GPL version 2.
Contact: m.futschik@biologie.hu-berlin.de
Supplementary information: Further information about the methods used in the OLIN software package can be found at http://itb.biologie.hu-berlin.de/~futschik/software/R/OLIN.