Efforts to provide centralized access in public databases for storing and disseminating large amounts of gene expression data have proved successful. The next challenge in data distribution will be providing normalized data across experiments to enable scientists to interrogate data in a larger scale across wider conditions. To achieve this goal, a standardized normalization algorithm suitable for both intra- and inter-experiment datasets is needed to ensure comparability. We developed a nonparametric normalization method called MDA (Maximal density approximation) by employing classification and approximation theory. The nonparametric feature imposes minimal requirement on data distributions, suggesting its applicability on cross-experiment data that tend to have higher degree of variability.
Our aim i s to extend the MDA method from intra- to inter-experiment data application and to derive an optimal standardized approach to meet the additional difficulties encountered in the analysis of inter-experiment data. Specifically, the proposed research aims first to establish a methodology for compatible inter-experiment data normalization. Extension of classification theory to image analysis will be addressed as well for improved performance. Finally, we propose to develop an algorithm enabling comparative analysis across incompatible datasets and technology platforms.
| Ni, Terri T; Lemon, William J; Shyr, Yu et al. (2008) Use of normalization methods for analysis of microarrays containing a high degree of gene effects. BMC Bioinformatics 9:505 |
