BrainGeneBot: a framework for variant prioritization and generative pretrained transformer-informed interpretation...

Area of Science:

• Genomics
• Computational Biology
• Neuroscience

Background:

• Polygenic risk scores (PRS) are crucial for Alzheimer's disease (AD) genetic susceptibility research.
• Dataset-specific biases in current PRS studies hinder variant prioritization, generalizability, and reproducibility.

Purpose of the Study:

• To develop a robust framework for prioritizing AD genetic risk variants by integrating multiple PRS datasets.
• To introduce BrainGeneBot, an AI tool for streamlining genomic analyses and enhancing PRS interpretability in AD research.

Main Methods:

• Proposed a transductive learning framework integrating multiple PRS datasets with genome-wide association study (GWAS) priority scores.
• Developed BrainGeneBot, an AI tool using generative pretrained transformers and retrieval-augmented generation for genomic analysis.
• Applied the framework to publicly available AD datasets from the PGS Catalog and compared it with unsupervised rank aggregation.

Main Results:

• The transductive learning approach validated known high-risk variants and identified novel ones correlating better with GWAS signals.
• The framework streamlined data retrieval and interpretation, enhancing genetic variant prioritization across multiple PRS studies.
• BrainGeneBot facilitated the discovery of biologically meaningful insights, improving PRS interpretability for AD research.

Conclusions:

• The proposed framework offers a robust approach for AD genetic research, improving data accessibility and accelerating discoveries.
• This method enhances the prioritization of genetic variants and provides deeper biological insights for AD.
• The AI-driven tool supports the development of precise AD interventions and treatments by refining genetic insights.