Based on Genetic Programming (GP), this paper proposed an approach called Best Rule Genetic Algorithm (BRGA) for optimizing classification rule, and gained the best classification rule set. This algorithm can adjust relevant parameters of classifier model, substantially improve the performance of classification by increasing appropriate iteration, therefore it has considerable flexibility and intelligibility. The performance of the proposed approach was evaluated by using six gene expression data sets through simulation. From the result, it is found that the proposed approach reduces computational complexity and redundancy with good classification accuracy and stability than approaches reported in other literatures so far.