An Integrated Approach of Support Vector Machine and Variable Neighborhood Search for Discovering Combinational Gene Signatures in Predicting Chemo-response of Osteosarcoma

摘要

Analyzing microarray data could help discover significant cancer genes and their mutual interactions, which can be used to generate hypothesis for the identification and validation of genetic biomarkers. However, the commonly used statistical significance analysis can only provide information of individual genes, thus neglecting influence of the mutual interactions. Therefore, methods aiming at discovering combinational gene signatures are highly valuable. In this paper, an integrated approach of support vector machine (SVM) and variable neighborhood search (VNS) is introduced in searching the gene signatures for predicting histologic response of chemotherapy on osteosarcoma patients. Cross validation results show that this method outperforms other existing algorithms. Further validation with the test dataset shows that only one of the fourteen samples is misclassified. The high testing accuracy further suggests that the proposed method has capability of extracting the stable discriminative signatures from microarray data.