ABSTRACT
Molecular proling has become a fundamental part of biomedical research, but much more importantly, it is now becoming a component of health care [1-3]. New clinical trials are identifying the benets of treatments supported by micro array data. This device for highly parallel measurements of gene expression was once restricted to pure research laboratories. It has now taken its place next to pathology reports and imaging devices as a tool to elaborate alterations in genetic pathways in order to provide clues to disease status and progression. It is this new paradigm that puts pressure on the mathematical algorithms and data manipulations that normalize molecular proling data. In order to obtain high precision, we must correct for inherent, repeatable biases in raw expression data. It is incumbent upon the biologist to understand, even anecdotally, the underlying principals of the proling devices themselves, from image acquisition to processing and data normalization. With this understanding, one knows whether data are being appropriately processed. In this
Abstract .................................................................................................................. 151 Introduction ............................................................................................................ 152 Background ............................................................................................................ 153 Technical Quality of Commercial Arrays ............................................................. 153 Expression Platform Types .................................................................................... 154 Alternative Expression Technologies ..................................................................... 156 Normalization Methods: History ........................................................................... 157 Normalization Methods for Commercial Arrays ................................................... 159 Causes and Consequences of Imprecise Measurements ........................................ 160 Cross-Normalization Comparisons-Experimental Design ................................. 161 Results .................................................................................................................... 163 Discussion .............................................................................................................. 166 Conclusion .............................................................................................................. 169 References .............................................................................................................. 170
chapter we describe several Affymetrix normalization methods along with a way to compare normalization methods using biologically interpretable analyses. We provide data that illuminates not only how normalization affects array-to-array precision, but also how one can use the gene ontology and gene regulatory and metabolic pathway tools to quickly interpret how normalization can affect the results.
