Abstract: Abstract: Purpose: Abnormalities always exist in breast epithelial tissues of normal histologic feature; thus, investigation of these differential expressed genes plays an important role in understanding of breast cancer development and in early diagnosis of this disease. Methods: Microarray technology provides a powerful tool to detect a large number of genes at the same point in time, which can be used to identify abnormal gene expression. In this study, we used bioinformatics tools to establish a model for early diagnosis of breast cancer, and screen differentially expressed genes by using signal pathway enrichment analysis. Results: The best prediction model was derived from the combination of differential genes enriched from KEGG and BioCarta database; the number of differential expressed genes in three random created prediction models was reduced from 22 to 7, 14 to 3 and 18 to 4, however, the prediction accuracy was consistently with the model established from all of the differentially expressed genes, and the average accuracy of all models is 96.3%. Conclusion: By this means, the prediction model can be simplified with the prediction accuracy unchanged, and thus facilitate the model apply to early diagnosis and prevention of breast cancer.

Key words: breast cancer, microarray, pathway, bioinformatics, prediction