Using Pre-training and Interaction Modeling for ancestry-specific disease prediction in UK Biobank

Area of Science:

• Genetics
• Genomic Epidemiology
• Computational Biology

Background:

• Genome-wide association studies (GWAS) reveal genetic underpinnings of complex traits but often lack diversity.
• Under-representation of non-European ancestries in genetic research creates a critical gap in disease prediction.
• Multiomic data offers potential to bridge this gap and improve health equity.

Conclusions:

• Multiomic data and advanced modeling techniques, including interaction terms and pre-training, can enhance disease prediction accuracy.
• Improvements were disease-specific and yielded moderate gains, highlighting the need for further research.
• Addressing ancestry gaps in genetic data is essential for equitable and effective disease prediction across populations.