基于XGBoost的质量性状基因互作检测方法  

A gene-based exchanged XGBoost method for detecting and ranking gene-gene interactions of qualitative trait

作　　者：郭颖婕[1] 李傲 刘晓燕[1] 郭茂祖[1,2] GUO Yingjie;LI Ao;LIU Xiaoyan;GUO Maozu(School of Computer Science and Technology,Harbin Institute of Technology,Harbin 150001,China;Beijing Key Laboratory of Intelligent Processing for Building Big Data,School of Electrical and Information Engineering,Beijing University of Civil Engineering and Architecture,Beijing 100044,China)

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001 [2]北京建筑大学电气与信息工程学院建筑大数据智能处理方法研究北京市重点实验室,北京100044

出　　处：《智能计算机与应用》2020年第3期202-208,共7页Intelligent Computer and Applications

基　　金：国家自然科学基金(61571163,61532014,61671189);国家重点研发计划项目(2016YFC0901902)。

摘　　要：在质量性状全基因组关联分析GWAS中,以基因作为研究单位的基因-基因相互作用检测方法,以其在统计效力与生物可解释性方面的优势备受关注。然而现有方法中多数对基因之间互作形式给出了强假设,降低了算法对互作关系的检测性能。针对已有方法存在的局限性,本文提出一种基于XGBoost的基因互作检测方法 geXGB。XGBoost作为一种流行且高效的机器学习方法,可以拟合基因型数据与表型之间的作用关系,并利用预测概率与加和模型之间的偏差表征相互作用关系的程度。geXGB对相互作用形式不作假设,增强该方法对不同形式相互作用的检测能力。仿真与真实实验结果表明:该方法能够有效进行不同类型相互作用的检测,可以应用于全基因组关联研究。Among the various statistical methods for identifying gene-gene interaction in qualitative genome-wide association studies(GWAS),gene-based methods have recently grown in popularity as they confer advantages in both statistical power and biological interpretability.However,most of these gene-based methods make strong assumptions on the form of the relationship between traits and SNPs,resulting in limited statistical power.The paper proposes a gene-based method based on XGBoost,a popular and highly effective method in machine learning,to model the relationship between genotype and traits,and then measure the interaction of gene pairs by the deviation of the predicted probability from a multiplicative model.This method makes fewer assumptions on the exact form of interaction,which may overcome some of the shortcomings in previous methods.In experiments with both simulation study on pure and strict disease models and real world data,the proposed method outperforms previous approaches in detecting interactions accurately.

关 键 词：XGBoost 基因相互作用 单核苷酸多态性位点 质量性状 全基因组关联分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]