Large Impact of Genetic Data Processing Steps on Stability and Reproducibility of Set-Based Analyses in Genome-Wide...

Area of Science:

• Human Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) are crucial for human genetics research.
• Reproducibility issues in GWAS are frequently observed, impacting result stability.
• Set-based analyses, common in hypothesis testing and effect size estimation, are particularly susceptible to these challenges.

Purpose of the Study:

• To analyze and address reproducibility challenges in set-based hypothesis testing and effect size estimation within GWAS.
• To investigate how minor data processing variations impact GWAS results.
• To provide practical tools for quantifying and mitigating data quality obstacles in GWAS.

Main Methods:

• Qualitative and quantitative description of data processing challenges, framed within a model misspecification context.
• Analytical calculation of power differences and bias arising from data cleaning choices in set-based analyses.
• Development and validation of tools for assessing and improving data quality in GWAS.

Main Results:

• Small, often overlooked, differences in data processing pipelines can lead to significant variations in GWAS results.
• Analytical calculations quantify the impact of data cleaning choices on statistical power and bias.
• Empirical analysis of a pancreatic cancer dataset demonstrated that key genetic associations can be lost due to data preparation inconsistencies.

Conclusions:

• Inconsistencies in data processing pipelines pose substantial threats to the robustness and reliability of GWAS findings.
• Explicit documentation and standardization of data processing steps are essential for reproducible GWAS.
• The developed tools aid researchers in identifying and overcoming data quality issues, enhancing the credibility of genetic association studies.