Abstract
This paper proposes a genetic algorithm to find the optimal array sets for microarray experimental design problems. Based on family competition, heterogeneous pairing selection and two new genetic operators, the proposed method can find the optimal designs of limited experimental materials under a statistical model (ANOVA). The proposed method is applied to several design problems whose numbers of target mRNA samples range from 5 to 70, which are more extensive than classical studies, with different number of arrays. We apply A-optimality criterion to get best possible designs with the smallest average variance when comparisons between all pairs of treatments are of equal interest. Experimental results demonstrate that our approach can find the optimum of each testing problem rapidly.