A novel gene set annotation optimization approach is proposed to improve the quality of gene annotations. Experimental results indicate that the proposed method effectively improves gene set annotation quality based on the GO structure and gene expression data.