Bayesian optimization algorithm-based methods searching for risk/protective factors

作     者：WEI Bin PENG QinKe CHEN Xiao ZHAO Jing 

作者机构：Engineering University of CAPF Electronic and Information School Xi'an Jiaotong University Key Laboratory for Intelligent Networks and Network Security of Ministry of Education Xi'an Jiaotong University State Key Laboratory for Manufacturing Systems Engineering and School of Electronic and Information Engineering Xi'an Jiaotong University 

出 版 物：《Chinese Science Bulletin》 (Chinese Science Bulletin)

年 卷 期：2013年第58卷第23期

页      面：2828-2835页

核心收录：

学科分类：1002[医学-临床医学] 08[工学] 1010[医学-医学技术（可授医学、理学学位）] 100215[医学-康复医学与理疗学] 10[医学] 081202[工学-计算机软件与理论] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported by the National Natural Science Foundation of China(60774086 and 61173111) Ph.D.Program Foundation of Ministry of Education of China(20090201110027) 

主　　题：贝叶斯算法 保护因子 风险 单核苷酸多态性 基础 贝叶斯优化算法 NP-hard 粒子群优化算法 

摘      要：The risks of developing complex diseases are likely to be determined by single nucleotide polymorphisms (SNPs), which are the most common form of DNA variations. Rapidly developing genotyping technologies have made it possible to assess the influence of SNPs on a particular disease. The aim of this paper is to identify the risk/protective factors of a disease, which are modeled as a subset of SNPs (with specified alleles) with the maximum odds ratio. On the basis of risk/protective factor and the relationship between nucleotides and amino acids, two novel risk/protective factors (called k-relaxed risk/protective factors and weighted-relaxed risk/protective factors) are proposed to consider more complex disease-associated SNPs. However, the enormous amount of possible SNPs interactions presents a mathematical and computational challenge. In this paper, we use the Bayesian Optimization Algorithm (BOA) to search for the risk/protective factors of a particular disease. Determining the Bayesian network (BN) structure is NP-hard; therefore, the binary particle swarm optimization was used to determine the BN structure. The proposed algorithm was tested on four datasets. Experimental results showed that the algorithm proposed in this paper is a promising method for discovering SNPs interactions that cause/prevent diseases.