The Cassa Lab recently published a new study in Genome Medicine, “From text to translation: using language models to prioritize variants for clinical review.” The study explores how artificial intelligence can help make better use of the large amount of clinical genetics information already contained in public databases such as ClinVar.
Many genetic variants identified through sequencing remain classified as “variants of uncertain significance,” meaning there is not yet enough evidence to determine whether they are likely to contribute to disease. In this work, the team developed ClinVar-BERT, a language model trained to analyze free-text variant summaries and identify patterns of evidence that may support pathogenic or benign interpretations. The model was trained using diagnostic lab reports and was evaluated using expert-curated variant classifications, functional screening data, and computational prediction scores.
The study found that language models can help prioritize variants that may benefit from expert re-review, including variants in clinically actionable genes. This work highlights how AI-based approaches may support clinical genetics by helping experts focus attention on variants where existing evidence may be especially informative. The paper was published on May 19, 2026, in Genome Medicine.