Imputed Gene Expression versus Single Nucleotide Polymorphism in Predicting Gray Matter Phenotypes.

Area of Science:

• Neuroscience
• Psychiatric Genetics
• Computational Biology

Background:

• Psychiatric disorders have a significant genetic component, prompting research into predicting traits from genetic data for early detection and personalized treatments.
• Genetically-regulated expression (GRE) captures tissue-specific genetic effects on gene expression, offering a novel approach for genetic association studies.
• Comparing GRE-based polygenic risk scores (gPRS) with traditional SNP-based polygenic risk scores (sPRS) is crucial for advancing psychiatric trait prediction.

Approach:

• Utilized UK Biobank data (34,149 individuals) to assess genetic prediction of 13 schizophrenia-related gray matter networks.
• Computed GRE for 56,348 genes across 13 brain tissues using MetaXcan and GTEx.
• Calculated gPRS and sPRS using SNP and gene effect sizes from training sets to predict brain phenotypes in testing sets.

Key Points:

• Both gPRS and sPRS successfully predicted brain phenotypes, with accuracy increasing with larger training sample sizes.
• gPRS demonstrated significantly higher prediction accuracies compared to sPRS across all 13 brain phenotypes.
• The predictive advantage of gPRS was more pronounced with training sample sizes below approximately 15,000.

Conclusions:

• GRE is a valuable genetic variable for brain phenotype association and prediction studies.
• gPRS shows superior performance over sPRS, particularly in scenarios with limited training data.
• Future imaging genetics studies should consider incorporating GRE for enhanced predictive power.