[論文レビュー] Statistical inference in two-sample summary-data Mendelian randomization using robust adjusted profile score

Qingyuan Zhao, Jingshu Wang+3 more|arXiv (Cornell University)|Jan 29, 2018

Genetic Associations and Epidemiology参考文献 49被引用数 258

ひとこと要約

本論文は、要約データを用いた二サンプルMRにおいて、プロファイルスコアを調整して多重効果に対処することで、頑健で一貫した推定を実現し、系統的および個別的な多重効果に対する頑健版（RAPS）を含む推定方法を開発した。

ABSTRACT

Mendelian randomization (MR) is a method of exploiting genetic variation to unbiasedly estimate a causal effect in presence of unmeasured confounding. MR is being widely used in epidemiology and other related areas of population science. In this paper, we study statistical inference in the increasingly popular two-sample summary-data MR design. We show a linear model for the observed associations approximately holds in a wide variety of settings when all the genetic variants satisfy the exclusion restriction assumption, or in genetic terms, when there is no pleiotropy. In this scenario, we derive a maximum profile likelihood estimator with provable consistency and asymptotic normality. However, through analyzing real datasets, we find strong evidence of both systematic and idiosyncratic pleiotropy in MR, echoing the omnigenic model of complex traits that is recently proposed in genetics. We model the systematic pleiotropy by a random effects model, where no genetic variant satisfies the exclusion restriction condition exactly. In this case we propose a consistent and asymptotically normal estimator by adjusting the profile score. We then tackle the idiosyncratic pleiotropy by robustifying the adjusted profile score. We demonstrate the robustness and efficiency of the proposed methods using several simulated and real datasets.

研究の動機と目的

• Motivate and formalize two-sample MR with GWAS summary data under potential pleiotropy and measurement error.
• Develop estimators for the causal effect that are consistent and asymptotically normal under no pleiotropy and under pleiotropy.
• Propose robust adjustments to the profile score to handle systematic pleiotropy and idiosyncratic pleiotropy.
• Assess estimator performance through simulations and real-data applications.
• Provide guidance on instrument strength, weak instruments, and selection bias in summary-data MR.

提案手法

• Derive the profile likelihood for the causal effect parameter beta by profiling out nuisance gamma parameters.
• Show consistency and asymptotic normality of the LIML-like estimator under Model 1 (no pleiotropy) and Assumptions on IV strength.
• Introduce an adjusted profile score to handle systematic pleiotropy (Model 2) and a robust version (RAPS) for outliers (Model 3).
• Replace unknown variance components with plug-in estimates from summary data to obtain standard errors.
• Discuss weak instrument bias and compare with IVW, highlighting dependence on average IV strength kappa.
• Provide practical guidance on LD clumping, instrument selection, and independence assumptions in summary-data MR.

実験結果

リサーチクエスチョン

• RQ1 Can a profile likelihood/score approach yield consistent and asymptotically normal estimates of the MR causal effect using summary data?
• RQ2 How do pleiotropy and measurement error affect inference in two-sample MR with summary data, and can adjusted/profile-robust methods mitigate these effects?
• RQ3 What estimators are robust to systematic and idiosyncratic pleiotropy in summary-data MR?
• RQ4 How does instrument strength (kappa) influence consistency, efficiency, and potential bias of MR estimators under these models?
• RQ5 What practical guidance emerges for SNP selection and LD considerations to avoid selection bias and weak instrument bias?

主な発見

• A LIML-like estimator from the profile likelihood is consistent and asymptotically normal under no pleiotropy (Model 1).
• Under systematic pleiotropy (Model 2), a consistent and asymptotically normal estimator is achievable via an adjusted profile score.
• Under combined systematic and idiosyncratic pleiotropy (Model 3), the Robust Adjusted Profile Score (RAPS) provides robustness to outliers while retaining efficiency.
• The analysis shows the IVW estimator can be biased when instruments are weak, and the proposed methods improve inference by accounting for measurement error in gamma and Gamma.
• Standard errors can be consistently estimated by plug-in estimates of V1 and V2 from summary data, enabling valid inference.
• Empirical demonstrations indicate robustness and efficiency advantages of the proposed RAPS method over standard approaches.