Screening Pathogenic Genes Of Breast Cancer Based On Association Rules Mining

With the increasing incidence and mortality of breast cancer while the age of the patients decreasing every year,it is urgent to propose and promote new clinical treatments which is more accurate than the traditional ones.In particular,with the awareness of later marriage and childbearing among women in the world,more and more women have been detected with breast cancer during pregnancy or within one year after giving birth,namely,pregnancy breast cancer.The special stage of its onset limits the way of clinical treatment of pregnancy breast cancer.Gene therapy is a novel treatment for breast cancer.This study focus on screening breast cancer genes more accurately and effectively from the massive genetic data.When using traditional clustering methods to analyze genetic data,it is impossible to set one gene in multiple classes,which is hard to reflect the interaction of biological functions between genes.In order to overcome this problem and find out the clinical target of breast cancer gene therapy,this study proposes a method combining differential analysis and association rule mining to screen for the breast cancer gene.The main work results are as follows:1.This paper proposes an association rule mining algorithm based on differential analysis to analyze the data from gene expression chips.The method applies a combination algorithm of moderate T-test and fold change to screen out the differentially expressed genes in pregnant and non-pregnant women.These differentially expressed genes and their corresponding PTM modifications were separately mined for association rules,and effective rules were screened.The genes corresponding to the two association rules were regarded as breast cancer related genes,and the breast cancer gene was screened through biological function analysis.2.A new method for classifying differentially expressed genes is presented based on samples' information.During the study,the expression levels of all differential genes in each sample were compared with their average expression levels,Based on the comparison results,samples with expression level higher than the average ones are marked as ‘up',samples with expression level significantly lower than average ones are marked as ‘down'.Genes marked as 'up' in the sample constitute the up transaction set,and the genes marked 'down' in each sample are extracted to form the down transaction set.This method breaks the traditional habit of dividing gene sets based on overall sample information.It compares each sample expression value with the average expression value one by one,divides the genes by sample markers.This method explores the interaction between genes at the biological level by reversely examining the similarity between genes within the sample.The breast cancer pathogenic genes obtained in this study clearly shows the difference between pregnant and non-pregnancy breast cancer.It is showed that,differentially expressed genes detected only in pregnant breast cancer were mostly differentially downregulated genes;genes detected only in non-pregnant breast cancer are mostly differentially up-regulated genes.This discovery is bound to promote a deep understanding of pregnant and non-pregnant breast cancer.The pathogenic genes of breast cancer obtained from this study has good biological interpretation and high credibility.They are highly coincident with known breast cancerrelated genes,and the bio-enrichment analysis results show that they are significantly enriched in key breast cancer pathways.Further,gene function annotations indicate that these genes mediate the regulation of breast cancer from different levels and processes.In summary,the breast cancer pathogenic genes screened in this study have clinical and medical significance,providing the theoretical support in gene-targeted therapy for pregnant and non-pregnant breast cancer.All in all,this work presents theoretical basis for the diagnosis,treatment,and prognosis of breast cancer.