Yan Wei-Li
School of Public Health, Xinjiang Medical University, Urumqi 830054, China.
Yi Chuan. 2008 May;30(5):543-9. doi: 10.3724/sp.j.1005.2008.00543.
Since the first genome-wide association study on human age-related macular degeneration was reported by Science journal in 2005, a series of genome-wide association studies have been published on human complex diseases or traits, such as obesity, type 2 diabetes, coronary artery disease, Alzheimer's disease and so on. The study of human genetics has recently undergone a dramatic transition which is called "the first wave of genome-wide association study". Some issues in statistical analysis of genome-wide association studies were reviewed by this paper. First, statistical analysis guidelines, methods and examples for genome-wide association studies of different designs, including unrelated case-control studies, population-based studies, and family-based association studies; second, multiple testing correction of P values, including Bonferroni correction, step-down Bonferroni correction, permutation correction, and the correction based on false discovery rate; third, population stratification and its effect on inference of genotype-phenotype associations. The False Positive Re-port Probability has been successfully applied in a recent genome-wide association study on coronary artery disease to con-trol the population stratification. Although genetic statistical methodology has been greatly developed in control of false positive associations caused by multiple testing or population stratification, it is still not sufficient to achieve the goal. Replicating genotype-phenotype associations is the only way to identify true association between genetic markers and common disease traits. The first wave of genome-wide association studies is producing an impressive list of unexpected associations between genes or chromosomal regions and a broad range of diseases. Traditional statistical techniques are adequate for the analysis and interpretation of these results. However, much more sophisticated methods of statistical analysis are likely to be required as we delve further into the genome in the search for networks of interacting gene variants, or interactions be-tween gene-gene networks and environmental factors. Finally, some useful links about statistical software for genome-wide association studies were provided.
自2005年《科学》杂志报道了第一项关于人类年龄相关性黄斑变性的全基因组关联研究以来，一系列关于人类复杂疾病或性状的全基因组关联研究相继发表，如肥胖症、2型糖尿病、冠状动脉疾病、阿尔茨海默病等。人类遗传学研究最近经历了一次重大转变，即所谓的“第一波全基因组关联研究”。本文对全基因组关联研究的一些统计分析问题进行了综述。第一，不同设计的全基因组关联研究的统计分析指南、方法和实例，包括非亲缘病例对照研究、基于人群的研究和基于家系的关联研究；第二，P值的多重检验校正，包括Bonferroni校正、逐步Bonferroni校正、置换校正以及基于错误发现率的校正；第三，群体分层及其对基因型-表型关联推断的影响。假阳性报告概率已成功应用于最近一项关于冠状动脉疾病的全基因组关联研究中，以控制群体分层。尽管在控制多重检验或群体分层导致的假阳性关联方面，遗传统计方法有了很大发展，但仍不足以实现目标。重复基因型-表型关联是识别遗传标记与常见疾病性状之间真正关联的唯一途径。第一波全基因组关联研究正在产生一系列令人印象深刻的、关于基因或染色体区域与广泛疾病之间意外关联的清单。传统统计技术足以对这些结果进行分析和解释。然而，随着我们进一步深入基因组，寻找相互作用基因变异网络，或基因-基因网络与环境因素之间的相互作用，可能需要更复杂的统计分析方法。最后，提供了一些关于全基因组关联研究统计软件的有用链接。
