Breast Cancer Prognosis Prediction Based On Multi-omics Dat

Cancer is a malignant disease,and its morbidity and mortality are increasing year by year.Prognostic analysis of cancer patients is beneficial to their prognosis risk and survival status,and then guides clinical treatment.In recent years,single-omics data has been unable to explain the complex causal relationship in pathology,so it has gradually become a development trend to study multi-omics data to distinguish complex cancer tissues.Based on the UCSC Xena official website,this paper selects three data about breast cancer omics,gene copy number,RNA gene expression,and RNA exon expression,and conducts a multi-omics joint analysis to predict whether the patient’s survival time exceeds five years.According to the different dimensional characteristics of the three omics data,firstly,the kendall correlation coefficient is used for the initial screening of exon expression,and most of the features that have no significant relationship with the response variable are screened out,and then the variance threshold filtering is performed on the three omics data respectively,to filter out the features with little change,and then use m RMR to perform filter-type feature selection,and filter out 50 features that are significantly related to the response variable,and then verify that 50 is the number of local optimal feature subsets.Finally,the data of these three optimal subsets were integrated,and four machine learning methods of support vector machine,Xgboost,Logistic regression,and randomforest were used to predict the prognosis with 5-fold cross-validation.By comparing the accuracy rate and AUC value of the model,it was finally found that the prediction effect of the support vector machine algorithm was better than the other three algorithms,and the multi-omics data of colorectal cancer from the same official website was used to verify that the model had better generalization ability.