(Note: Current Rosetta Inpharmatics employees are shown in boldface type.)

Eric E. Schadt ^{1 4 *}, Cheng Li ³, Byron Ellis ², Wing H. Wong ^{2 3 *}

¹Department of Biomathematics, University of California, Los Angeles, California
² Department of Statistics, Harvard University, Cambridge, Massachusetts
³ Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts
4 Departments of Informatics, Rosetta Inpharmatics, Kirkland, Washington
email: Eric E. Schadt (eschadt@rosetta.org) Wing H. Wong (wwong@stat.harvard.edu)

*Correspondence to Eric E. Schadt, Department of Biomathematics, University of California, Los Angeles, CA and Departments of Informatics, Rosetta Inpharmatics, Kirkland, WA.

*Correspondence to Wing H. Wong, Departments of Biostatistics, Harvard School of Public Health, 655 Hungtington, Ave, Boston, MA.

Abstract

Algorithms for performing feature extraction and normalization on high-density oligonucleotide gene expression arrays, have not been fully explored, and the impact these algorithms have on the downstream analysis is not well understood. Advances in such low-level analysis methods are essential to increase the sensitivity and specificity of detecting whether genes are present and/or differentially expressed. We have developed and implemented a number of algorithms for the analysis of expression array data in a software application, the DNA-Chip Analyzer (dChip). In this report, we describe the algorithms for feature extraction and normalization, and present validation data and comparison results with some of the algorithms currently in use.