基于Relief和SVM-RFE的组合式SNP特征选择

针对SNP的全基因组关联分析面临SNP数据的高维小样本特性和遗传疾病病理的复杂性两大难点,将特征选择引入SNP全基因组关联分析中,提出基于Relief和SVM-RFE的组合式SNP特征选择方法。该方法包括两个阶段:Filter阶段,使用Relief算法剔除无关SNPs;Wrapper阶段,使用基于支持向量机的特征递归消减方法(SVM-RFE)筛选出与遗传疾病相关的关键SNPs。实验表明,该方法具有明显优于单独使用SVM-RFE算法的性能,优于单独使用Relief-SVM算法的分类准确率,为SNP全基因组关联分析提供了一种有效途径。

Combined SNP feature selection based on relief and SVM-RFE

The genome-wide association study (GWAS)on SNPs faces two big issues: high dimensional SNP data with small sample characteristics and complex mechanisms of genetic diseases. This paper proposed a combined SNP feature selection method through bring feature selection methods into GWAS. The method included two stages: filter stage, it used Relief algorithm to eliminate irrelevant SNP features, wrapper stage, it used support vector machine based recursive feature reduction (SVM-RFE) algorithm to select the key SNPs set. Experiments show that the proposed method has an obviously better performance than SVM RFE algorithm, and also gains higher classification accuracy than Relief-SVM algorithm, which provides an effective way for SNP genome-wide association analysis.