Evans David M, Davey Smith George
University of Queensland Diamantina Institute, Translational Research Institute, Brisbane, Queensland 4102, Australia; email:
Annu Rev Genomics Hum Genet. 2015;16:327-50. doi: 10.1146/annurev-genom-090314-050016. Epub 2015 Apr 22.
Mendelian randomization (MR) is an approach that uses genetic variants associated with a modifiable exposure or biological intermediate to estimate the causal relationship between these variables and a medically relevant outcome. Although it was initially developed to examine the relationship between modifiable exposures/biomarkers and disease, its use has expanded to encompass applications in molecular epidemiology, systems biology, pharmacogenomics, and many other areas. The purpose of this review is to introduce MR, the principles behind the approach, and its limitations. We consider some of the new applications of the methodology, including informing drug development, and comment on some promising extensions, including two-step, two-sample, and bidirectional MR. We show how these new methods can be combined to efficiently examine causality in complex biological networks and provide a new framework to data mine high-dimensional studies as we transition into the age of hypothesis-free causality.
孟德尔随机化（MR）是一种利用与可改变的暴露因素或生物中间体相关的基因变异来估计这些变量与医学相关结局之间因果关系的方法。尽管它最初是为研究可改变的暴露因素/生物标志物与疾病之间的关系而开发的，但其应用范围已扩展到分子流行病学、系统生物学、药物基因组学及许多其他领域。本综述的目的是介绍MR、该方法背后的原理及其局限性。我们考虑了该方法的一些新应用，包括为药物开发提供信息，并对一些有前景的扩展方法进行评论，包括两步法、两样本法和双向MR。我们展示了如何将这些新方法结合起来，以有效地检验复杂生物网络中的因果关系，并提供一个新框架来挖掘高维研究数据，从而步入无假设因果关系的时代。
