Zhu Huanhuan, Zhou Xiang
Department of Biostatistics, University of Michigan, Ann Arbor, MI 48109, USA.
Center for Statistical Genetics, University of Michigan, Ann Arbor, MI 48109, USA.
Quant Biol. 2021 Jun;9(2):107-121. doi: 10.1007/s40484-020-0207-4.
Genome-wide association studies (GWASs) have identified thousands of genetic variants that are associated with many complex traits. However, their biological mechanisms remain largely unknown. Transcriptome-wide association studies (TWAS) have been recently proposed as an invaluable tool for investigating the potential gene regulatory mechanisms underlying variant-trait associations. Specifically, TWAS integrate GWAS with expression mapping studies based on a common set of variants and aim to identify genes whose GReX is associated with the phenotype. Various methods have been developed for performing TWAS and/or similar integrative analysis. Each such method has a different modeling assumption and many were initially developed to answer different biological questions. Consequently, it is not straightforward to understand their modeling property from a theoretical perspective.
We present a technical review on thirteen TWAS methods. Importantly, we show that these methods can all be viewed as two-sample Mendelian randomization (MR) analysis, which has been widely applied in GWASs for examining the causal effects of exposure on outcome. Viewing different TWAS methods from an MR perspective provides us a unique angle for understanding their benefits and pitfalls. We systematically introduce the MR analysis framework, explain how features of the GWAS and expression data influence the adaptation of MR for TWAS, and re-interpret the modeling assumptions made in different TWAS methods from an MR angle. We finally describe future directions for TWAS methodology development.
We hope that this review would serve as a useful reference for both methodologists who develop TWAS methods and practitioners who perform TWAS analysis.
全基因组关联研究（GWAS）已鉴定出数千种与许多复杂性状相关的遗传变异。然而，它们的生物学机制在很大程度上仍然未知。转录组全关联研究（TWAS）最近被提出作为一种宝贵的工具，用于研究变异-性状关联背后潜在的基因调控机制。具体而言，TWAS将GWAS与基于一组共同变异的表达图谱研究相结合，旨在识别其基因调控表达（GReX）与表型相关的基因。已经开发出各种方法来进行TWAS和/或类似的综合分析。每种此类方法都有不同的建模假设，并且许多方法最初是为回答不同的生物学问题而开发的。因此，从理论角度理解它们的建模特性并非易事。
我们对13种TWAS方法进行了技术综述。重要的是，我们表明这些方法都可以被视为两样本孟德尔随机化（MR）分析，该分析已广泛应用于GWAS中以检验暴露对结局的因果效应。从MR角度看待不同的TWAS方法为我们理解它们的优点和缺陷提供了一个独特的视角。我们系统地介绍了MR分析框架，解释了GWAS和表达数据的特征如何影响MR在TWAS中的应用，并从MR角度重新解释了不同TWAS方法中所做的建模假设。我们最后描述了TWAS方法学发展的未来方向。
我们希望这篇综述能为开发TWAS方法的方法学家和进行TWAS分析的从业者提供有用的参考。
