Cedars-Sinai launches AI model for gene variant predictions

Researchers at Cedars-Sinai have created a new artificial intelligence model called DYNA. This model helps doctors identify harmful gene changes that can lead to diseases. It aims to improve the accuracy of disease diagnosis and support personalized medicine. In a study published in Nature Machine Intelligence, the team showed that DYNA performs better than current AI models in linking DNA mutations to specific health issues, particularly cardiovascular diseases. Dr. Huixin Zhan, one of the study's authors, noted that many genetic variants are unclear in their significance. DYNA addresses this uncertainty effectively. Most existing AI models can tell which genetic variants may harm protein functions. However, they struggle to connect those variants to particular diseases. DYNA successfully makes these connections, which is crucial for patient care. To build DYNA, researchers used a type of AI called a Siamese neural network to improve two models. They tested DYNA’s predictions against a public database called ClinVar, which tracks genetic variations. The results showed that DYNA accurately matched gene variants to the diseases they cause. Dr. Jason Moore, another study author, stated that DYNA offers a flexible framework for studying genetic diseases. Future enhancements could make it an essential tool for doctors to customize treatment based on a patient’s genetic information. The DYNA code is publicly available on GitHub.