Wang Huanhuan, Wu Xiang
IEEE/ACM Trans Comput Biol Bioinform. 2023 Jan-Feb;20(1):455-464. doi: 10.1109/TCBB.2022.3155774. Epub 2023 Feb 3.
Analyzing K-order Single Nucleotide Polymorphism (SNP) interactions through the statistics of Genome-Wide Association Studies (GWAS) is crucial for discovering pathogenic causes of human complex diseases and controlling risk genetic variants of diverse disorders. We propose a method based on Ant Colony Optimization (ACO) algorithm to detect gene interactions for GWAS - an Intelligent Privacy-Preserving scheme (IPP). Initially, we design a multi-objective search algorithm to discover the candidate SNP sets related to disease phenotype, which utilizes Differential Privacy (DP) method by disturbing the multi-objective function to construct a rational epistatic privacy protection strategy. Furthermore, the global path selection strategy composed of two probabilistic methods is proposed to reduce the probability of falling into the local optimum. We use simulated models and a real dataset of Rheumatoid Arthritis (RA) to compare IPP with four popular methods to detect K-order SNPs, the experimental results show that IPP can guarantee the search accuracy effectively and enhance the detecting ability of various models. Further, the privacy budget experiments indicate that the range of privacy budget in IPP is reasonable and make the framework more stable.
通过全基因组关联研究（GWAS）的统计数据来分析K阶单核苷酸多态性（SNP）相互作用，对于发现人类复杂疾病的致病原因以及控制各种疾病的风险基因变异至关重要。我们提出了一种基于蚁群优化（ACO）算法的方法来检测GWAS中的基因相互作用——一种智能隐私保护方案（IPP）。首先，我们设计了一种多目标搜索算法来发现与疾病表型相关的候选SNP集，该算法通过干扰多目标函数利用差分隐私（DP）方法构建合理的上位性隐私保护策略。此外，还提出了一种由两种概率方法组成的全局路径选择策略，以降低陷入局部最优的概率。我们使用模拟模型和类风湿性关节炎（RA）的真实数据集，将IPP与四种流行的检测K阶SNP的方法进行比较，实验结果表明IPP能够有效地保证搜索准确性，并提高各种模型的检测能力。此外，隐私预算实验表明IPP中的隐私预算范围是合理的，并且使框架更加稳定。
