We recently read with great interest the paper by Yin and colleagues [1] that pharmacologically available genomic data, cis-eQTL/cis-pQTL from human blood and skeletal muscle tissue, and GWAS pooled data on sarcopenia related traits were used to analyse the potential causal relationship between drug target genes and sarcopenia. The study employed colocalization and Mendelian randomization (MR) analyses to identify 17 potential therapeutic targets for sarcopenia. However, in the articles by Yin et al. [2], they put forward their own views on this method of analysis. Colocalization analysis after MR analysis may introduce uncorrelated pleiotropic effects due to violation of the exclusion–restriction assumption, which cannot strengthen the MR results.

The thriving development of genome-wide association studies (GWAS) over the past decades has laid the foundation for the explosive growth of MR studies in recent years. Recently, an increasing number of MR studies have begun to explore the efficacy and safety of novel drugs and to seek new applications for existing drugs [3]. Colocalization methods were invented to discover how disease-associated genetic variants revealed by GWAS affect downstream pathways and whether genetic variants may share one or more causal variants with potential biological mediators; they are commonly used as sensitivity analyses after MR analysis to strengthen the reliability of the results [4]. Identifying and classifying pleiotropy is a crucial step in MR research, which can be divided into vertical pleiotropy and horizontal pleiotropy, with horizontal pleiotropy further divided into uncorrelated horizontal pleiotropy and correlated horizontal pleiotropy. In MR studies, when the exposure of genetic instrumental variable (IV) affects the outcome through its impact on downstream traits, it is vertical pleiotropy, representing the essence of the MR methodology [5]. Moreover, when exposure IV not only influences the outcome through exposure but also affects the outcome through common confounding factors between exposure and outcome, it is referred to as correlated horizontal pleiotropy. However, uncorrelated horizontal pleiotropy occurs when other genetic variations (such as single nucleotide polymorphisms, SNPs), due to linkage disequilibrium (LD), collectively impact the outcome alongside the exposure to genetic IV [6].

Although uncorrelated horizontal pleiotropy can be addressed by methods such as MR-Egger or MR-PRESSO, it may still introduce potential biases to MR results. In cis-MR, colocalization between expression quantitative trait locus (eQTL) and GWAS data suggests the sharing of a genetic locus between upstream genes and downstream traits, indicating vertical pleiotropy [4]. This strengthens the causality in MR and contradicts Yin's claim of violating the third assumption. However, in polygenic MR analysis [7], the colocalization of GWAS with GWAS data indicates the sharing of a genetic locus between two phenotypes. This implies the existence of uncorrelated horizontal pleiotropy, violating the exclusion–restriction assumption, thus undermining support for MR results. After detecting horizontal pleiotropy through the MR-Egger intercept test, the MR pleiotropy residual sum and outlier (MR-PRESSO) test, or other methods, the CAUSE method can be used to comprehensively consider horizontal pleiotropy [6]. It can determine whether the pleiotropy between exposure and outcome is uncorrelated horizontal pleiotropy or correlated horizontal pleiotropy, thus cautiously interpreting the MR results.

Employing the CAUSE methodology enables researchers to interpret MR outcomes with increased caution, identifying the specific type of pleiotropy influencing the linkage between exposure and outcome. This enhancement in precision and trustworthiness of MR investigations is particularly significant amidst intricate biological pathways and genetic mechanisms. Therefore, the CAUSE method provides a powerful tool for MR analysis, which allows for a comprehensive consideration of horizontal pleiotropy when detected, thus inferring causal relationships more accurately. MR and colocalization indeed occupy critical roles within the fields of genetic epidemiology and genomics research, and the result of MR is typically interpreted as the causal relationships between exposure and outcome; colocalization results are used to unveil genetic architectures and biological mechanisms [4]. Despite differences in concept and practice between MR and colocalization analysis, both approaches leverage genetic variations to investigate the relationships among traits, serving as critical tools in contemporary genetic epidemiology.

Lately, with the publication of high-quality GWAS and eQTL data, the transcriptome-wide association studies (TWAS) approach has been extensively utilized to uncover the connections between gene expression levels and complex traits. TWAS employs eQTL data to build predictive models of gene expression, which are then utilized to evaluate the relationship between gene expression levels and traits. Through this method, researchers can indirectly assess the impact of gene expression on traits without directly measuring gene expression levels. The majority of trait-associated genes identified through TWAS are physically well separated from other candidate genes; thus, they are less influenced by LD than those identified by GWAS [8]. Cis-MR and TWAS are both methods that utilize genetic variations to study the association between gene expression and traits. They share conceptual similarities, particularly in using genetic variations to infer potential causal relationships between gene expression and traits. The advancement of TWAS research provides new tools and algorithms for MR studies. When conducting MR research, we can use the latest TWAS algorithms such as MR-JTI [9] and cTWAS [10] to support our MR results. These algorithms offer a robust approach to enhance our causal inference regarding the association between gene expression levels and traits. By integrating the use of these methods, researchers can validate their findings from various angles, thereby enhancing the credibility and precision of causal inference.

Therefore, based on the above discussion, we recommend that in conducting MR analysis, if conducting cis-MR studies, leveraging colocalization findings can bolster conclusions, supplemented by TWAS analysis to enhance the reliability of MR results. Whereas in polygenic MR analysis, we should cautiously interpret colocalization results, employ multiple analytical methods for cross-validation, such as CAUSE methods, and conduct analysis as per the requirements of STROBE-MR guidelines.

我们最近饶有兴趣地阅读了 Yin 及其同事的论文 [1]，该论文使用了药理学可用的基因组数据、来自人类血液和骨骼肌组织的 cis-eQTL/cis-pQTL 以及关于肌肉减少症相关特征的 GWAS 汇总数据来分析药物靶基因与肌肉减少症之间的潜在因果关系。该研究采用共定位和孟德尔随机化 （MR） 分析来确定肌肉减少症的 17 个潜在治疗靶点。然而，在 Yin 等人 [2] 的文章中，他们对这种分析方法提出了自己的观点。由于违反了排除-限制假设，MR 分析后的共定位分析可能会引入不相关的多效性效应，这不能增强 MR 结果。

过去几十年全基因组关联研究 （GWAS） 的蓬勃发展为近年来 MR 研究的爆炸式增长奠定了基础。最近，越来越多的 MR 研究开始探索新药的有效性和安全性，并为现有药物寻求新的应用 [3]。发明了共定位方法，以发现 GWAS 揭示的疾病相关遗传变异如何影响下游途径，以及遗传变异是否可能与潜在的生物介质共享一个或多个因果变异;它们通常用作 MR 分析后的敏感性分析，以增强结果的可靠性 [4]。识别和分类多效性是 MR 研究中的关键步骤，可分为垂直多效性和水平多效性，水平多效性进一步分为不相关的水平多效性和相关的水平多效性。在 MR 研究中，当遗传工具变量 （IV） 的暴露通过其对下游性状的影响影响结果时，它是垂直多效性的，代表了 MR 方法的本质 [5]。此外，当暴露 IV 不仅通过暴露影响结果，而且还通过暴露和结果之间的共同混杂因素影响结果时，它被称为相关的水平多效性。然而，当由于连锁不平衡 （LD） 导致的其他遗传变异（如单核苷酸多态性，SNP）与遗传 IV 暴露一起共同影响结果时，就会发生不相关的水平多效性 [6]。

尽管不相关的水平多效性可以通过 MR-Egger 或 MR-PRESSO 等方法解决，但它仍可能给 MR 结果带来潜在的偏倚。在 cis-MR 中，表达数量性状位点 （eQTL） 和 GWAS 数据之间的共定位表明上游基因和下游性状之间共享一个遗传位点，表明垂直多效性 [4]。这加强了 MR 中的因果关系，并与 Yin 声称违反第三个假设的说法相矛盾。然而，在多基因 MR 分析 [7] 中，GWAS 与 GWAS 数据的共定位表明两种表型之间共享一个遗传位点。这意味着存在不相关的水平多效性，违反了排除限制假设，从而削弱了对 MR 结果的支持。通过 MR-Egger 截距试验、MR 多效残差和异常值 （MR-PRESSO） 检验或其他方法检测水平多效性后，CAUSE 方法可用于综合考虑水平多效性 [6]。它可以确定暴露和结果之间的多效性是不相关的水平多效性还是相关的水平多效性，从而谨慎解释 MR 结果。

采用 CAUSE 方法使研究人员能够更加谨慎地解释 MR 结果，确定影响暴露与结果之间联系的特定多效性类型。在错综复杂的生物途径和遗传机制中，MR 研究精度和可信度的提高尤为重要。因此，CAUSE 方法为 MR 分析提供了一个强大的工具，可以在检测到时综合考虑水平多效性，从而更准确地推断因果关系。MR 和共定位确实在遗传流行病学和基因组学研究领域中占据关键地位，MR 的结果通常被解释为暴露与结果之间的因果关系;共定位结果用于揭示遗传结构和生物学机制 [4]。尽管 MR 和共定位分析在概念和实践上存在差异，但这两种方法都利用遗传变异来研究性状之间的关系，是当代遗传流行病学的关键工具。

最近，随着高质量 GWAS 和 eQTL 数据的发表，全转录组关联研究 （TWAS） 方法被广泛用于揭示基因表达水平与复杂性状之间的联系。TWAS 利用 eQTL 数据构建基因表达预测模型，然后用于评估基因表达水平和性状之间的关系。通过这种方法，研究人员可以间接评估基因表达对性状的影响，而无需直接测量基因表达水平。通过 TWAS 鉴定的大多数性状相关基因在物理上与其他候选基因很好地分离;因此，与 GWAS 识别的 LS 相比，它们受 LD 的影响更小 [8]。Cis-MR 和 TWAS 都是利用遗传变异来研究基因表达与性状之间关联的方法。它们在概念上具有相似性，特别是在使用遗传变异来推断基因表达和性状之间的潜在因果关系方面。TWAS 研究的进步为 MR 研究提供了新的工具和算法。在进行 MR 研究时，我们可以使用最新的 TWAS 算法，如 MR-JTI [9] 和 cTWAS [10] 来支持我们的 MR 结果。这些算法提供了一种强大的方法来增强我们对基因表达水平和性状之间关联的因果推断。通过整合这些方法的使用，研究人员可以从各个角度验证他们的发现，从而提高因果推理的可信度和准确性。

因此，基于上述讨论，我们建议在进行 MR 分析时，如果进行 cis-MR 研究，利用共定位结果可以支持结论，并辅以 TWAS 分析以提高 MR 结果的可靠性。而在多基因 MR 分析中，我们应该谨慎解释共定位结果，采用多种分析方法进行交叉验证，例如 CAUSE 方法，并按照 STROBE-MR 指南的要求进行分析。