School of Computer & Information Engineering, Nanning Normal University, Nanning, 530001, China.
School of Computer science, Fudan University, Shanghai, 200433, China.
BMC Genomics. 2021 Dec 20;22(Suppl 1):910. doi: 10.1186/s12864-021-08207-8.
Identification of epistatic interactions provides a systematic way for exploring associations among different single nucleotide polymorphism (SNP) and complex diseases. Although considerable progress has been made in epistasis detection, efficiently and accurately identifying epistatic interactions remains a challenge due to the intensive growth of measuring SNP combinations.
In this work, we formulate the detection of epistatic interactions by a combinational optimization problem, and propose a novel evolutionary-based framework, called GEP-EpiSeeker, to detect epistatic interactions using Gene Expression Programming. In GEP-EpiSeeker, we propose several tailor-made chromosome rules to describe SNP combinations, and incorporate Bayesian network-based fitness evaluation into the evolution of tailor-made chromosomes to find suspected SNP combinations, and adopt the Chi-square test to identify optimal solutions from suspected SNP combinations. Moreover, to improve the convergence and accuracy of the algorithm, we design two genetic operators with multiple and adjacent mutations and an adaptive genetic manipulation method with fuzzy control to efficiently manipulate the evolution of tailor-made chromosomes. We compared GEP-EpiSeeker with state-of-the-art methods including BEAM, BOOST, AntEpiSeeker, MACOED, and EACO in terms of power, recall, precision and F1-score on the GWAS datasets of 12 DME disease models and 10 DNME disease models. Our experimental results show that GEP-EpiSeeker outperforms comparative methods.
Here we presented a novel method named GEP-EpiSeeker, based on the Gene Expression Programming algorithm, to identify epistatic interactions in Genome-wide Association Studies. The results indicate that GEP-EpiSeeker could be a promising alternative to the existing methods in epistasis detection and will provide a new way for accurately identifying epistasis.
确定上位性相互作用为探索不同单核苷酸多态性（SNP）和复杂疾病之间的关联提供了一种系统的方法。尽管在检测上位性方面已经取得了相当大的进展，但由于 SNP 组合的大量增加，有效地、准确地识别上位性相互作用仍然是一个挑战。
在这项工作中，我们将检测上位性相互作用形式化为组合优化问题，并提出了一种新的基于进化的框架，称为 GEP-EpiSeeker，使用基因表达编程来检测上位性相互作用。在 GEP-EpiSeeker 中，我们提出了几种定制的染色体规则来描述 SNP 组合，并将基于贝叶斯网络的适应度评估纳入定制染色体的进化中，以发现可疑的 SNP 组合，并采用卡方检验从可疑 SNP 组合中识别最佳解决方案。此外，为了提高算法的收敛性和准确性，我们设计了两个具有多个和相邻突变的遗传操作符，以及一个具有模糊控制的自适应遗传操作方法，以有效地操纵定制染色体的进化。我们在 12 个 DME 疾病模型和 10 个 DNME 疾病模型的 GWAS 数据集上，将 GEP-EpiSeeker 与 BEAM、BOOST、AntEpiSeeker、MACOED 和 EACO 等最先进的方法进行了比较，从功效、召回率、精度和 F1 得分方面进行了比较。我们的实验结果表明，GEP-EpiSeeker 优于比较方法。
在这里，我们提出了一种新的方法，名为 GEP-EpiSeeker，它基于基因表达编程算法，用于识别全基因组关联研究中的上位性相互作用。结果表明，GEP-EpiSeeker 可能是检测上位性的现有方法的一种有前途的替代方法，并将为准确识别上位性提供一种新的方法。
