AlphaGenome-enabled analysis of non-coding regulatory variants underlying RHD expression with wet-lab validation

Area of Science:

• Human genetics
• Genomics
• Bioinformatics

Background:

• Identifying functional non-coding regulatory variants is crucial but challenging in human genetics.
• Conventional methods like CRISPR screening and GWAS are costly and time-consuming.
• The RHD antigen's expression is not fully explained by coding variants alone, necessitating investigation of non-coding regions.

Purpose of the Study:

• To develop and validate an integrated AI-guided and experimental framework for identifying functional non-coding regulatory variants impacting RHD expression.
• To demonstrate the utility of AI, specifically AlphaGenome, in prioritizing regulatory variants within the RHD locus.
• To provide a scalable and cost-effective alternative to traditional screening methods for non-coding variant analysis.

Main Methods:

• Applied AlphaGenome, a deep-learning model, to systematically analyze the RHD locus using multi-omics data.
• Utilized in silico deletion and SNP perturbation analyses to predict variant effects on RHD expression.
• Performed CRISPR-mediated base editing at prioritized non-coding SNP sites in K562 cells, followed by qPCR and flow cytometry for validation.

Main Results:

• AlphaGenome prioritized regulatory regions in the promoter, 5'UTR, and intragenic regions of the RHD locus.
• In silico analyses predicted suppressive effects of variants in promoter and intragenic regions.
• Experimental validation showed strong concordance between AI predictions and observed RHD transcriptional and phenotypic outcomes, with targeted base editing of high-priority variants demonstrating significant functional effects.

Conclusions:

• The combination of AI-guided variant prioritization and base editing is a powerful, scalable, and cost-effective strategy for identifying functional non-coding variants.
• This framework has direct implications for genomics-based RHD typing and advancing transfusion medicine.
• This study represents the first phenotypic validation of AlphaGenome predictions using wet-lab experiments.