一种分析全基因组上位性的新方法

发布时间：2018-09-18 06:54

【摘要】：传统基于单位点的全基因组关联研究存在重复性低、难以解释性等缺陷,而采用基于机器学习的上位性分析中面临计算复杂度高、预测准确度不足等问题.本文提出一种分析全基因组上位性的新方法,该方法采用二阶段框架的上位性分析方法,它包含特征过滤阶段以及上位性组合优化阶段,在特征过滤阶段提出了多准则融合策略,从多个不同角度评价遗传变异位点,以保证易感的弱效位点能被保留,然后采用多准测排序融合策略剔除与疾病状态关联程度低的遗传变异,进一步在上位性组合优化阶段采用贪婪算法启发式地搜索组合空间,以降低时间复杂度,最后采用支持向量机作为上位性评价模型.实验中采用不同的连锁不平衡参数与经典算法SNPruler与ACO的性能进行对比,实验结果表明:本文方法能有效保留弱效位点,一定程度上提高了疾病预测的正确度.
[Abstract]:The traditional genome association research based on unit points has some shortcomings such as low repeatability and difficult to explain. However, the epistatic analysis based on machine learning is faced with the problems of high computational complexity and low prediction accuracy. In this paper, a new method is proposed to analyze the epistasis of the whole genome. This method uses the two-stage framework of epistasis analysis, which includes the stage of feature filtering and the phase of epistatic combination optimization. In the stage of feature filtering, a multi-criteria fusion strategy is proposed to evaluate the genetic variation sites from different angles to ensure that the susceptible weak sites can be preserved. Then the multi-quasi-test ranking fusion strategy is adopted to eliminate the genetic variation with low correlation with disease state, and then the greedy algorithm is used to heuristically search the combination space in the stage of epistatic combinatorial optimization, so as to reduce the time complexity. Finally, support vector machine is used as the epistatic evaluation model. In the experiment, different linkage disequilibrium parameters are compared with the performance of the classical algorithms SNPruler and ACO. The experimental results show that the proposed method can preserve the weak sites effectively and improve the accuracy of disease prediction to a certain extent.
【作者单位】：湖南大学信息科学与工程学院;湖南工学院计算机科学与信息学院;
【基金】：国家自然科学基金资助项目(61672223) 湖南省自然科学基金资助项目(2016JJ4029)~~
【分类号】：Q811.4;TP18

【相似文献】