Abstract: Self-Organizing Maps（SOM） and the hierarchical clustering are two of the most classical clustering technologies for analyzing gene expression data,which exist own advantages and disadvantages on account of the complexity and the instability of gene expression data.Therefore,on base of comparing difference of the two clustering technologies,this article creatively proposes one new algorithm,that is first clustering gene expression data with SOM and second clustering the weight of nerve cells corresponding the clustering from the first step.In succession,the new algorithm is applied to the published data of yeast gene expression to prove that it conquers some bug of SOM and improves the efficiency of gene clustering through emulation mode.