Normalization of cDNA Microarray Data Using Wavelet Regressions
Abstract:Normalization is an essential step in microarray data mining and analysis. For cDNA microarray data, the primary purpose of normalization is removing the intensity-dependent bias across different slides within an experimental group or between multiple groups. The locally weighted regression (lowess) procedure has been widely used for this purpose but can be comparatively time consuming when the dataset becomes relatively large. In this study, we applied wavelet regressions, a new smoothing method for recovering a regression function from data that is supposed to outperform other methods in many cases, such as spline or local polynomial fitting, to normalize two cDNA microarray datasets. Relative to the lowess procedure, we found that wavelet regressions not only produced reliable normalization results but also ran much faster. The computing speed represents one of the most important advantages over other algorithms, especially when one is interested in analyzing a large microarray experiment involving hundreds of slides.
Document Type: Review Article
Affiliations: Mail Code 7792, 7703 Floyd C Drive, San Antonio, TX 78229, USA.
Publication date: 2004-12-01
More about this publication?
- Combinatorial Chemistry & High Throughput Screening publishes full length original research articles and reviews describing various topics in combinatorial chemistry (e.g. small molecules, peptide, nucleic acid or phage display libraries) and/or high throughput screening (e.g. developmental, practical or theoretical). Ancillary subjects of key importance, such as robotics and informatics, will also be covered by the journal. In these respective subject areas, Combinatorial Chemistry & High Throughput Screening is intended to function as the most comprehensive and up-to-date medium available. The journal should be of value to individuals engaged in the process of drug discoveryand development, in the settings of industry, academia or government.