Microarray experiments are characterized by the presence of many sources of experimental bias and a remarkably large technical variability. The assessment of differential expression for genes transcribed into a small number of mRNA copies heavily depends on the proper quantification of background fluorescence within spot. The rough model ‘observed = hybridization plus background’ fluorescence is at first reformulated at spot level, then it is embedded into a Bayesian hierarchical model suited for fitting control spots. The novelties of the approach include the background correction performed on the latent mean of replicated spots, and an explicit model for outlying observations at low fluorescence values in which the probability of occurrence and their magnitude depend on the background fluorescence intensity. The analysis of unpublished data from a maize ear tissues experiment confirms the feasibility of MCMC inferences as regard the computational burden. A Bayesian hierarchical model for spot fluorescence in microarrays Federico M. Stefanini Department of Statistics ‘G.Parenti’ University of Florence, Italy March 27, 2007 viale Morgagni 59, 50134 Firenze, Italy, stefanin@ds.unifi.it Hosted by The Berkeley Electronic Press Abstract Microarray experiments are characterized by the presence of many sources of experimental bias and a remarkably large technical variability. The assessment of differential expression for genes transcribed into a small number of mRNA copies heavily depends on the proper quantification of background fluorescence within spot. The rough model ‘observed = hybridization plus background’ fluorescence is at first reformulated at spot level, then it is embedded into a Bayesian hierarchical model suited for fitting control spots. The novelties of the approach include the background correction performed on the latent mean of replicated spots, and an explicit model for outlying observations at low fluorescence values in which the probability of occurrence and their magnitude depend on the background fluorescence intensity. The analysis of unpublished data from a maize ear tissues experiment confirms the feasibility of MCMC inferences as regard the computational burden.
[1]
Raymond J Carroll,et al.
DNA Microarray Experiments: Biological and Technological Aspects
,
2002,
Biometrics.
[2]
Philip Heidelberger,et al.
Simulation Run Length Control in the Presence of an Initial Transient
,
1983,
Oper. Res..
[3]
T. Grisar,et al.
Housekeeping genes as internal standards: use and limits.
,
1999,
Journal of biotechnology.
[4]
Walter R. Gilks,et al.
Model checking and model improvement
,
1995
.
[5]
B. Korn,et al.
Normalization of array hybridization experiments in differential gene expression analysis.
,
1999,
Nucleic acids research.
[6]
John Geweke,et al.
Evaluating the accuracy of sampling-based approaches to the calculation of posterior moments
,
1991
.
[7]
Sylvia Richardson,et al.
Markov Chain Monte Carlo in Practice
,
1997
.
[8]
E. Wolski,et al.
Normalization strategies for cDNA microarrays.
,
2000,
Nucleic acids research.
[9]
Adrian E. Raftery,et al.
[Practical Markov Chain Monte Carlo]: Comment: One Long Run with Diagnostics: Implementation Strategies for Markov Chain Monte Carlo
,
1992
.