Stochastic approach to engineering and analysis of DNA-based microarray technology.

摘要

Microarray assays are powerful tools for the investigation of cellular activity at the genome-scale. They exploit the hybridization reactions among nucleic acid polymers to quantify gene expression (mRNA) levels. However, in recent years, several studies have shown that different types of cDNA microarray assays yield different results for the quantification of differences in gene expression. This phenomenon is anecdotally imputed to cross-hybridization: the undesired hybridization of non-target cDNA molecules to microarray probes. However, until now characterization of the identities of cross-hybrids to a microarray probe and the amounts of cross-hybridizing cDNA species has not been possible. The goal of this dissertation is to characterize the interaction of probes and cDNA molecules in microarray assays at the dynamic level, and analyze and improve the reliability of cDNA microarray technology. To this end, a stochastic approach to the chemical kinetics is employed to investigate the hybridization reaction network underlying the microarray experiments. First, we consider the stochastic approach in a general sense using the stochastic master equation, exploring autocatalysis and the simple hybridization between one cDNA molecule and one oligonucleotide probe. Next, the stochastic approach is used to develop a fast and efficient stochastic simulation algorithm for the simulation of genome-level microarray assays. Using this tool, a robust method is developed to characterize probe-cDNA interactions in DNA microarray assays, combining stochastic simulation, robust parametric estimation of hybridization thermodynamics, and the methodologies of hypothesis testing. This technique permits evaluation of the sensitivity and specificity of each probe as well as the specific cross-hybridization patterns. The technique is used to evaluate the reliability of a commercial Agilent array for S. Cerevisiae genome, as well as three probe sets that are generated by popular oligomer design tools: OligoArray 2, OligoWiz 2, and PICKY 2. Finally, a novel method of designing oligonucleotide microarray probe sets is developed based on judicious combination of sets designed using other oligomer design algorithms. The strategy is applied to develop a set of reliable probes from OligoArray 2, OligoWiz 2, and PICKY 2 for the genome of S. Cerevisiae that is superior to the probe sets from which it is constructed.