Establishment Of Breast Cancer Prognostic Model Based On Conditionally Correlated Complementary Genes

At present,the proportion of women suffering from breast cancer in the world is increasing,which seriously affects women’s physical and mental health.Although the diagnosis and treatment of breast cancer have improved in the past few years,nearly half of breast cancer patients will still die due to recurrence,making the survival rate of breast cancer patients not effectively improved.Therefore,in order to improve the clinical treatment effect of breast cancer patients and reduce their mortality,it is necessary to study and establish a prognostic model of breast cancer from the molecular mechanism.Aiming at the characteristics of breast cancer gene expression data with few samples but many genes and the complex relationship between genes,a screening method for grouping conditionally related complementary genes is proposed,and then each group is subjected to stepwise Cox regression to establish the final Prognosis model of breast cancer.Firstly,the data set is preprocessed and differentially analyzed,and genes that are significantly differentially expressed in cancer samples are selected to save time for feature screening;secondly,the conditionally associated complementary genes in the significantly differentially expressed genes are screened and grouped,and each Group genes were respectively used in the training set using stepwise Cox regression fitting,verified in the test set,and a breast cancer prognosis model containing 10 genes was determined.Among them,VWCE,SPDYC,CRYBG3,DEFB1,SEL1L2,NMNAT2 are unfavorable factors for survival rate,and 4 genes,AMZ1,GJB2,CXCL2,and ALDOC,are favorable factors for survival rate.By establishing a Bayesian network structure for these 10 genes,an important gene module was also identified.In order to test the effectiveness of the prognostic model,the risk scores of breast cancer patients in the training set,test set,and overall data set were calculated,and the samples of the three data sets were divided into high and low risks according to the median of the risk scores in the training set Group,draw K-M survival curves in turn.K-M survival analysis shows that the survival curves of the two groups are significantly different,and the P values are all less than 0.05 after passing the log-rank test.Secondly,in order to evaluate the accuracy of these 10 gene prognostic models,The model evaluation index C-index value is calculated to be equal to 0.81,and the time-dependent ROC curves of 5 years and 10 years are drawn for the training set and the test set respectively,and the final calculated AUC value can reach 0.7 or more.The experimental results show that the method of establishing a breast cancer prognosis model based on associated complementary genes can reduce the dimensionality of high-dimensional data,eliminate the problem of collinearity between genes,and improve the speed and accuracy of model building.The model is robust and interpretable,and the prognostic model established by 10 genes can help patients in clinical prediction.