Abstract P2-05-37: In-situ hybridization of microRNA and support vector machines–based prognostic classifiers for breast cancer

Purpose: We plan to use a tissue microarray-based microRNA(miRNA) detection by in situ hybridization with LNA probe and support vector machine on base of data mining to construct the Prognostic model for breast cancer and triple negative breast cancer (TNBC). Methods: Combining the BreastMark miRNA database and papers published, we chose candidate miRNAs. Then a tissue microarray-based miRNA detection in situ hybridization with LNA probe was used to detect miRNA expression in 445 breast cancer tissue. Univariate analysis with Kaplan-Meier identified independent prognostic factors, 445 patients was divided into the training set and validation set (70% VS 30%) by software R(version 3.2.3), then support vector machine on base of data mining was used to construct the Prognostic model for breast cancer and triple negative breast cancer via software R and predict.svm(e1071). Results: 15 miRNAs were detected from 445 breast cancer tissue and univariate analysis with Kaplan-Meier identified significant prognostic miRNAs: miR-361-5p, miR-301a, miR-223, miR-421, miR-454, miR-493 were the independent prognostic factors for DFS. By using data mining technique, we combined clinical factors(stage, tumor size, lymphnode status, molecular subtype) and miRNAs mentioned above to establish the personalized predictive mathematical models SVM-BC and SVM-TNBC. In the training set and validation set of SVM-BC model, the AUC value of ROC curve for DFS were 0.86,0.83. The total accuracy,sensitivity,specificity of SVM-BC model for DFS were 94%,73%,99% and 84%,57%,92% in the training set and validation set. The SVM-BC model can divide breast cancer patients into high-risk and low-risk groups, and 5-year survival rates for low-risk patients were higher than that for high-risk patients(96.8% VS 39.3%, p In the training set and validation set of SVM-TNBC model, the AUC value of ROC curve for DFS were 0.84,0.86. The total accuracy,sensitivity,specificity of SVM-TNBC model for DFS were 90%,71%,97% and 84%,55%,94% in the training set and validation set. The SVM-TNBC model can divide breast cancer patients into high-risk and low-risk groups, and 5-year survival rates for low-risk patients were higher than that for high-risk patients(93.9 % VS 23.5%, p Conclusion: The Prognostic model basing of tissue microarray-based miRNA detection in situ hybridization with LNA probe and support vector machine for prognosis can effectively predict the prognosis of breast cancer and TNBC. Citation Format: Cao Z, Yao L, Hu X, Zhang Z, Shao Z. In-situ hybridization of microRNA and support vector machines–based prognostic classifiers for breast cancer [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P2-05-37.