Unfolding of Microarray Data

The use of DNA microarrays for the analysis of complex biological samples is becoming a mainstream part of biomedical research. One of the most commonly used methods compares the relative abundance of mRNA in two different samples by probing a single DNA microarray simultaneously. The simplicity of this concept sometimes masks the complexity of capturing and processing microarray data. On the basis of the analysis of many of our microarray experiments, we identified the major causes of distortion of the microarray data and the sources of noise. In this study, we provide a systematic statistical approach for extraction of true expression ratios from raw microarray data, which we describe as an unfolding process. The results of this analysis are presented in the form of a model describing the relationship between the measured fluorescent intensities and the concentrations of mRNA transcripts. We developed and tested several algorithms for inference of the model parameters for the microarray data. Special emphasis is given to the statistical robustness of these algorithms, in particular resistance to outliers. We also provide methods for measurement of noise and reproducibility of the microarray experiments.