R.O. Dror. Noise models in gene array analysis. Area exam report, MIT Department of Electrical Engineering and Computer Science, 2001.
[Full text, PDF]

Abstract

Gene arrays measure the expression levels of thousands of genes simultaneously, providing an extremely powerful tool for biology and medicine. Unfortunately, the high level of noise in the resulting measurements often obscures the biological processes of interest. After providing an overview of gene array technology, we consider several recent expression level estimation methods which deal with this noise explicitly. We also consider the implications of estimation methods based on noise models in higher-level processing of gene array data.

This report focuses on the work of Li and Wong (PNAS, 2001), Friedman et al. (JCB, 2000), and Hughes et al. (Cell, 2000). It includes a rederivation of the "Rosetta error model" described by Hughes et al., including an examination of the assumptions underlying this model.