基于迭代回归的基因相互作用定义方法  

Define genetic interaction based on iterative regression

作　　者：邓甦[1] 姜杨 付长贺 DENG Su;JIANG Yang;FU Changhe(College of Mathematics and Systems Science, Shenyang Normal University, Shenyang 110034, China;School of Statistics and Mathematics, Inner Mongolia University of Finance and Economics, Hohhot 010070, China)

机构地区：[1]沈阳师范大学数学与系统科学学院,沈阳110034 [2]内蒙古财经大学统计与数学学院,呼和浩特010070

出　　处：《沈阳师范大学学报（自然科学版）》2021年第5期419-422,共4页Journal of Shenyang Normal University:Natural Science Edition

基　　金：智能计算与信息处理教育部重点实验室课题(2020ICIP05);辽宁省科技厅自然科学基金资助项目(2019-MS-285);国家自然科学基金资助项目(61703290)。

摘　　要：基因相互作用是生物信息学中的重要课题,主要研究在非等位基因缺失条件下基因之间的功能联系,为细胞信号通路分析和药物重定位等提供了新的方向。经典的基因相互作用定义方式采用确定的预期中立函数,但是这种固定的中立函数往往不能反映具体实验数据的特征,而基于机器学习定义基因相互作用的方法可以较好地解决这个问题。首先,通过回归分析方法拟合符合数据特征的预期中立函数;其次,剔除数据集中残差比较大的点,在剩余数据中重复这一过程,直到回归系数趋于收敛;最后,将得到的回归方程作为预期中立函数定义所有样本的基因相互作用。在包含660多万条记录的酵母菌数据集上的实验结果表明,基于迭代回归的定义方式能够发现更精准的相互作用。Gene interaction is an important subject in bioinformatics.It mainly studies the functional relationship between genes under the condition of non-allelic deletion,which provides a new direction for cell signal pathway analysis and drug relocation.The classical way to define gene interaction is to use a certain expected neutral function,but this fixed neutral function cannot reflect the characteristics of the specific experimental data.This paper proposes a method to define gene interaction based on machine learning.Firstly,it fit the expected neutral function in accordance with the data characteristics by regression analysis,and then the points with large residuals in the data set are eliminated,and the process is repeated in the remaining data until the regression coefficient tends to converge.Finally,as the expected neutral function,the received regression equation is used to define gene interactions for all samples.The experimental results based on the yeast data set containing more than 6.6 million records show that more comprehensive interaction can be found by the definition method based on iterative regression,which are uniform and more consistent with the biological law.

关 键 词：回归 机器学习 预期中立函数 基因相互作用 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] Q348[自动化与计算机技术—控制科学与工程]