When pipelines run but coordinates fail: A simple spatial specificity check for false locality in post-GWAS analysis.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Post-Genome-Wide Association Studies (GWAS) analysis software is crucial for identifying gene-disease associations.
• Software failures can lead to biologically invalid results, termed 'false locality', where findings appear localized but are based on incorrect genomic data.

Purpose of the Study:

• To identify mechanisms leading to false locality in post-GWAS analysis.
• To propose a validation method ensuring the biological validity and spatial specificity of analysis results.

Main Methods:

• Investigated three mechanisms of false locality: genome-build address mismatch, non-specific regional noise, and location-label blindness.
• Conducted an audit of 91 sentinel coordinates and an SMR/HEIDI CXCL10 test with shifted labels.
• Proposed a 'Change Test' to validate spatial specificity by altering analysis window or label.

Main Results:

• Genome-build mismatches (GRCh38 vs. GRCh37) caused significant coordinate shifts (>100kb for 66, >200kb for 54 of 91 tested).
• Non-specific noise and location-label blindness were demonstrated, where results remained significant or identical despite incorrect genomic data or labels.
• The proposed Change Test effectively identified false locality by requiring output changes upon intentional data manipulation.

Conclusions:

• False locality is a significant issue in post-GWAS software, leading to biologically invalid findings.
• The proposed Change Test provides a crucial validation step, transforming software execution from a passive check to an active spatial-specificity verification.