Abstract: We proposed an improved parallel immune genetic algorithm. The introduced island concept on the population size can make the sizes of populations variable; The immune operator introduced in the fitness functioncan avoid the algorithm premature convergence. Hence, this method can prevent local convergence in the optimal algorithm in a great extent, and make the probability of approaching the global convergence bigger. The algorithm was used via the Yeast Saccharomyces cerevisiae cell cycle gene expression profile from SGD to coexpress clusters. Compared to the experiment analyzed by Spellman for functional genomics, the efficiencyof this algorithm in the functional genomics can be proved.

Key words: dynamic fuzzy clustering, parallel immune genetic algo rithm, immune operator, island