Illumina Upgrades Genomic AI to Decode Rare Disease Variants

SpliceAI2 functions by predicting the impact of splicing—the critical process where RNA transcripts are assembled before protein creation. When these transcripts are disrupted, they can lead to severe hereditary diseases. Traditional identification methods often struggle to flag these cryptic splice sites, leaving researchers with variants of uncertain significance. Illumina’s BioInsight AI Lab trained the new model on a dataset 100 times larger than the original version, resulting in a 34% improvement in splice site usage quantification compared to existing alternatives.

This release expands Illumina’s suite of genomic tools, which now includes PromoterAI and PrimateAI-3D. Together, these models allow scientists to scan the genome for splice, promoter, and missense variants with double the efficacy of previous iterations. According to the company, SpliceAI2 results show closer alignment with downstream protein effects observed in UK Biobank data, providing a more reliable bridge between DNA sequencing and clinical outcomes. The tool is now integrated into Illumina's DRAGEN and Emedgene platforms, offering researchers a streamlined path to prioritize the most impactful genetic signals in their datasets.