The goal of this training proposal is to design, implement, and validate a mixed effects model extension of the S-score algorithm (Zhang et al., J Mol Biol, 2001) for oligonucleotide microarrays. The S-score was originally developed to provide alternatives to existing software for measuring differential gene expression. It is based on an error model in which the detected signal is proportional to the probe pair signal for highly expressed genes, but approaches a background level (rather than 0) for low levels of expression. This model is used to calculate a relative change in probe pair intensities that converts probe signals into multiple measurements with equalized errors, which are summed to form the S-score. Validation studies confirmed that the S-score outperformed many other methods. However, improvements on the S-score may be realized by extending it to a more general model capable of handling more than two samples and mixed effects in the predictor variables. The use of a mixed effects model more closely describes microarray studies, where certain factors represent a subset of the population being studied. Such a model captures the correlation structure of microarray experiments more accurately and offers greater power in detecting gene expression changes. Under his mentors, the PI will develop a mixed effects model extension and corresponding software algorithms in the R language. This will lead to the creation and widespread distribution of a software algorithm incorporating the latest innovations in gene expression analysis.
| Kennedy, Richard E; Archer, Kellie J; Miles, Michael F (2006) Empirical validation of the S-Score algorithm in the analysis of gene expression data. BMC Bioinformatics 7:154 |
