A Critical Evaluation of Background Gene Omission in Imaging Transcriptomics

Area of Science:

• Neuroscience
• Genomics
• Bioinformatics

Background:

• Imaging transcriptomics links brain imaging phenotypes to gene expression.
• Overrepresentation analysis (ORA) is commonly used to interpret gene sets related to imaging-derived phenotypes (IDPs).
• Previous studies often neglect appropriate background gene selection for ORA.

Purpose of the Study:

• To evaluate the impact of background gene selection on ORA results in imaging transcriptomics.
• To assess the prevalence of proper background reporting in published imaging transcriptomics studies.
• To provide guidance on selecting appropriate backgrounds for ORA.

Main Methods:

• Systematic review of 152 imaging transcriptomics studies (2015-2024).
• Simulations to assess the effect of background size on ORA significance.
• Comparison of ORA results using different background gene sets (e.g., all protein-coding genes vs. AHBA genes).

Main Results:

• 84.9% of reviewed studies failed to report background genes for ORA.
• Only 5.26% of studies utilized the Allen Human Brain Atlas (AHBA) gene set as background.
• Default backgrounds significantly inflated pathway significance (up to 50-fold), while AHBA backgrounds maintained significance near 0.05.

Conclusions:

• Inappropriate background selection inflates ORA findings in imaging transcriptomics, particularly for synaptic and neurotransmission pathways.
• The use of the AHBA gene set provides a more reliable background for ORA.
• Transparent reporting of background gene selection is essential for the validity of imaging transcriptomics research.