Deep learning and statistical methods identify novel asthma risk variants in Europeans

Area of Science:

• Genetics
• Respiratory Medicine
• Bioinformatics

Background:

• Asthma is a common heritable respiratory disease with a complex genetic background.
• Genome-wide association studies (GWAS) have identified numerous asthma risk loci, but the complete genetic architecture is not yet understood.

Purpose of the Study:

• To enhance the understanding of asthma's genetic landscape in individuals of European ancestry.
• To improve polygenic risk prediction for asthma using advanced statistical and deep learning methodologies.

Main Methods:

• Conducted the largest GWAS meta-analysis for asthma in European ancestry populations, integrating data from two major initiatives.
• Employed pleiotropy-informed multi-trait analysis (MTAG) and conditional false discovery rate (condFDR) with eosinophil counts.
• Utilized a Transformer-based deep learning framework for variant prioritization and polygenic risk score (PRS) development.

Main Results:

• Identified 69 novel independent genome-wide significant loci associated with asthma.
• MTAG, condFDR, and deep learning approaches revealed additional candidate loci.
• Functional analysis implicated new genes in immune regulation, airway remodeling, and metabolism; deep learning-derived PRS models showed superior performance.

Conclusions:

• Generated a comprehensive map of asthma-associated loci in European populations.
• Significantly improved polygenic risk prediction for asthma.
• Provided a foundation for future mechanistic research into asthma pathogenesis.