Conditional Inference in Cis-Mendelian Randomization Using Weak Genetic Factors

Abstract

Mendelian randomization (MR) is a widely used method to estimate the causal effect of an exposure on an outcome by using genetic variants as instrumental variables. MR analyses that use variants from only a single genetic region (cis‐MR) encoding the protein target of a drug are able to provide supporting evidence for drug target validation. This paper proposes methods for cis‐MR inference that use many correlated variants to make robust inferences even in situations, where those variants have only weak effects on the exposure. In particular, we exploit the highly structured nature of genetic correlations in single gene regions to reduce the dimension of genetic variants using factor analysis. These genetic factors are then used as instrumental variables to construct tests for the causal effect of interest. Since these factors may often be weakly associated with the exposure, size distortions of standard t‐tests can be severe. Therefore, we consider two approaches based on conditional testing. First, we extend results of commonly‐used identification‐robust tests for the setting where estimated factors are used as instruments. Second, we propose a test which appropriately adjusts for first‐stage screening of genetic factors based on their relevance. Our empirical results provide genetic evidence to validate cholesterol‐lowering drug targets aimed at preventing coronary heart disease.

Unraveling the role of plasma proteins in dementia: insights from two cohort studies in the UK, with causal evidence from Mendelian randomization

“…Next, we used a two-sample MR study design, based predominantly on genetic variants located in or near genes that encode the relevant drug targets, to infer causality from protein concentration → dementia (cis-MR). [81–84] Cis-MR is considered to be less susceptible to pleiotropy, and the potential effect of a drug by analyzing the genomic locus encoding protein targets, which may be informative for drug trial design. [82, 84]…”

“…[81–84] Cis-MR is considered to be less susceptible to pleiotropy, and the potential effect of a drug by analyzing the genomic locus encoding protein targets, which may be informative for drug trial design. [82, 84]…”

“…Next, we used a two-sample MR study design, based predominantly on genetic variants located in or near genes that encode the relevant drug targets, to infer causality from protein concentration → dementia (cis-MR). [81][82][83][84] Cis-MR is considered to be less susceptible to pleiotropy, and the potential effect of a drug by analyzing the genomic locus encoding protein targets, which may be informative for drug trial design. [82,84] The selection of the primary cis-MR models included SNPs successfully harmonized with the gene encoding regions, with the flanking region being within 10-kilobase (kB) in either direction of the start and stop coordinates for the genes, according to Human Genome reference release GRCh38.…”

“…[81][82][83][84] Cis-MR is considered to be less susceptible to pleiotropy, and the potential effect of a drug by analyzing the genomic locus encoding protein targets, which may be informative for drug trial design. [82,84] The selection of the primary cis-MR models included SNPs successfully harmonized with the gene encoding regions, with the flanking region being within 10-kilobase (kB) in either direction of the start and stop coordinates for the genes, according to Human Genome reference release GRCh38. LD clumping was then done to remove the excess of most highly correlated variants at each locus, with R 2 threshold set at < 0.001, retaining the SNPs with the strongest associations with the protein of interest filtered by P < 5 × 10 −5 .…”

simmr: An open-source tool to perform simulations in Mendelian Randomization

“…We mentioned in the Introduction section that at least 45 statistical methods for performing Mendelian Randomization (MR) exist. These methods include IVW (Burgess et al, 2013), dIVW (Ye et al, 2021), pIVW (Xu et al, 2022), MR-Egger (Bowden et al, 2016), MR-RAPS (Zhao et al, 2020b), MRAID (Yuan et al, 2022), MRMix (Qi and Chatterjee, 2019), MR-cML (Xue et al, 2021), MVMR-cML (Lin et al, 2023), MR-PRESSO (Verbanck et al, 2018), IMRP (Zhu et al, 2021), MR-Median (Bowden et al, 2016), MR-MaxLike (Burgess et al, 2016b), MR-Corr (Cheng et al, 2022a), MR-Robust (Rees et al, 2019), MR-Lasso (Kang et al, 2016), MR-Conmix (Burgess et al, 2020), CAUSE (Morrison et al, 2020), MR-CUE (Cheng et al, 2022b), MR-Horse (Grant and Burgess, 2023), MR-BMA (Zuber et al, 2020), MR-Robin (Gleason et al, 2020), EMIC (Jiang et al, 2022a), MR-Mode (Hartwig et al, 2017), MRBEE (Lorincz-Comi et al, 2023), MR-Lap (Mounier and Kutalik, 2023), the Wald test (Palmer et al, 2008), JAM (Newcombe et al, 2016), MR using factor analysis (Patel et al, 2023), mixIE (Lin et al, 2021), MRMO (Deng et al, 2022), BMRMO (Deng et al, 2023), BWMR (Zhao et al, 2020a), moPMR-Egger (Liu et al, 2021), sisVIVE (Kang et al, 2016), MR-LDP (Cheng et al, 2020), MR-CIP (Xu et al, 2021), MR-PATH (Iong et al, 2020), MR-Clust (Foley et al, 2021), BayesMR (Bucur et al, 2020), BMRE (Schmidt and Dudbridge, 2018), MR-link (van Der Graaf et al, 2020), OMR (Wang et al, 2021b), CoJo (Yang et al, 2012), MR using PCA (Burgess et al, 2017), and GRAPPLE (Wang et al, 2021a).…”