Basic Science and Pathogenesis.

Area of Science:

• Genetics
• Neuroscience
• Bioinformatics

Background:

• Investigating rare genetic variants in Alzheimer's disease (AD) is difficult due to limited functional prediction specificity in brain cells.
• Developed brain cell-type-specific non-coding variant prediction scores incorporating splicing and enhancer/promoter effects.
• Introduced a novel rare variant testing method applied to whole-genome sequencing (WGS) data from the Alzheimer's Disease Sequencing Project (ADSP).

Purpose of the Study:

• To enhance the identification of rare genetic variants associated with Alzheimer's disease.
• To improve the specificity of functional predictions for non-coding variants in specific brain cell types.
• To uncover novel AD-associated genes and regulatory elements using advanced analytical tools.

Main Methods:

• Trained deep learning models on microglia RNA-seq data to predict variant effects on splicing and gene regulation.
• Developed a splicing modifier score (SMS) and a Bayesian probabilistic model for rare variant discovery and functional annotation weighting.
• Applied these methods to WGS data from 7,966 AD cases and 13,412 controls.

Main Results:

• Models achieved high performance in predicting microglia splicing (PR-AUC 0.853) and fine-mapped sQTL variants (ROC-AUC 0.656).
• MPRA validation confirmed functional effects of AD-associated variants.
• Identified 10 novel genetic associations for AD, with splicing annotations showing strong enrichment for non-coding rare variants.

Conclusions:

• Introduced a comprehensive framework for brain cell-type-specific rare variant analysis in AD.
• The framework enhances genome-wide association tests, improving the identification of AD-relevant genes and regulatory elements.
• Uncovered previously unidentified rare variant associations in AD, offering potential targets for future therapeutic development.