Alphagenome-enabled analysis of non-coding regulatory variants underlying rhd expression

Abstract Introduction RhD is a critical blood group antigen associated with hemolytic transfusion reactions and fetal/newborn disease. While coding variants have been extensively studied through Sanger sequencing and SNP arrays, non-coding variants remain poorly understood. Whole-genome sequencing (WGS) and CRISPR methods are costly and time-consuming. Although functional variants in the RhD promoter, 5'UTR, and introns have been reported, systematic validation is lacking. We established cell models and integrated conventional analyses with AlphaGenome (an AI tool launched in June 2025) to map RhD regulatory regions and key non-coding variants, demonstrating the transformative potential of AI-experimental integration in blood group antigen studies. Methods We developed molecular techniques for RhD mRNA expression and isoform analysis in K562 and HuDEP2 cells. Custom PCR and qPCR assays targeted full-length RhD/RhCE transcripts and splice isoforms, with RNA-seq data informing assay design. Flow cytometry detected surface RhD without RhCE cross-reactivity. Whole-genome sequencing (WGS) profiled SNPs/mutations across the RhD locus. AlphaGenome played a pivotal role in regulatory region prediction, refining the vast number of genomic regions to be analyzed into critical regulatory sites (e.g., promoter, 5'UTR, and intron-exon junctions) by integrating histone marks (UCSC) and other epigenetic data, significantly enhancing analysis efficiency and accuracy. Results K562 cells exhibited strong RhD/RhCE expression, while HUDEP2 cells showed high RhD but low RhCE expression. Three of 11 RhD isoforms were detected. AlphaGenome precisely identified Chr1:25272393-25272547 (promoter/5’UTR) and intron-exon junctions as critical regulatory regions, predicting 217 variants. GnomAD revealed 26 variants with frequencies below 0.001%. Clinically validated variants (c.1–110A>C, c.1-83C>T) were located within AlphaGenome-predicted regions. Histone predictions substantially overlapped with AlphaGenome at the promoter/5’UTR, confirming its predictive reliability. K562 WGS identified 84 variants (6 upstream, 5 exonic, 65 intronic, 8 downstream), none of which were clinically significant. Conclusion We developed experimental systems and pioneered the use of AlphaGenome for studying RhD non-coding blood group variants. AlphaGenome's ability to refine regulatory region predictions significantly enhanced research efficiency, providing an innovative tool for non-coding variant analysis. These molecular, cellular, and bioinformatic approaches, combined with future CRISPR editing, support genomics-based RhD typing. Integration of transcriptome and protein data further enhances mechanistic insights. These protocols are applicable to erythroid cell lines and adaptable to other clinically significant blood group genes, offering a scalable framework for blood group antigen research.