Design and quality control of large-scale two-sample Mendelian randomization studies

Area of Science:

• Genetics
• Epidemiology
• Bioinformatics

Background:

• Mendelian randomization (MR) studies are prone to metadata and analytical errors, potentially biasing results.
• The Fatty Acids in Cancer Mendelian Randomization Collaboration (FAMRC) aimed to address these issues.

Purpose of the Study:

• To develop and implement a quality control (QC) pipeline for MR studies.
• To identify and correct metadata and analytical errors in fatty acid and cancer GWAS data.

Main Methods:

• Collated summary association statistics from 6 fatty acid and 54 cancer genome-wide association studies (GWAS).
• Applied a comprehensive QC pipeline comparing study data to external references (NHGRI-EBI GWAS Catalogue, 1000 Genomes).
• Identified errors by comparing reported vs. expected genetic effect sizes using GWAS hits and a reference panel.

Main Results:

• Identified metadata errors or analytical issues in 7 out of 54 studies (11.6%).
• Resolved errors to create a dataset of 219,842 genetic associations across 90 cancer types.
• The dataset involved 566,665 cancer cases and 1,622,374 controls.

Conclusions:

• A significant proportion of studies (11.6%) contained errors that could bias MR analyses.
• The developed QC pipeline enhances the integrity of summary data, increasing the reliability of downstream MR findings.
• The QC pipeline is publicly available as the CheckSumStats R package for broader research use.