An Evolutionary Algorithm for Optimization of cDNA Microarray Experimental Designs

Microarrays constitute a powerful tool for practical livestock applications including, amongothers, diagnostics, target identification, screening, and genotyping. But they are also costly,both in time and resources, which makes the careful design of microarray experiments criticalto generate useful data cost effectively. Statistical analysis of data generated from welldesigned experiments allows for meaningful biological correlation. As with any otherexperimental approach, to succeed, the objectives of the study must be clearly stated. Thisneed has tended to be brushed aside since the quantity of data generated falsely suggests thatalmost any possible question can be addressed (Simon et al. 2002). Regrettably this is not so,and with elevated costs and high demands on time it has become clear that microarray studieshave to be well defined as to their objectives and well planned to ensure that the questions ofinterest can be effectively addressed. The planning stage encompasses experimental design,which is the focus of this paper. To find the best overall design that adequately balancesconflicting constraints is not a trivial task. Microarray experimental design is essentially amulticriteria optimization problem. For this class of problems Evolutionary Algorithms arewell suited for they can search the multicriteria solution space and evolve a design thatoptimizes the parameters of interest based on their relative value to the researcher under agiven set of constraints. This paper introduces the use of Genetic Algorithms (GAs), a class ofEvolutionary Algorithms, for optimization of experimental designs of spotted microarraysusing a weighted multicriteria objective function.