What AI model, developed by the Debora Marks' Lab at Harvard, can identify genetic variants most likely to cause severe disease and death?

The AI model developed by Debora Marks' Lab at Harvard, known as PopEVE, is a groundbreaking tool designed to pinpoint genetic variants most likely to cause severe disease and even death. This innovative artificial intelligence model is capable of scoring every possible amino acid substitution across all human proteins, offering a comprehensive and calibrated view of a patient's genome. Its primary goal is to help clinicians more quickly and accurately diagnose rare, single-variant genetic diseases, significantly shortening the diagnostic journey for many patients.

PopEVE achieves this remarkable capability by building upon prior research and integrating vast amounts of data. It combines deep evolutionary information, gleaned from hundreds of thousands of different species, with human population genetic variation data from large repositories like the UK Biobank. This unique blend allows PopEVE to discern which parts of the roughly 20,000 human proteins are essential for life and which can tolerate changes. By understanding these evolutionary constraints and human population patterns, the model can not only identify disease-causing mutations but also rank their severity, distinguishing between benign variants and those likely to lead to severe conditions or even childhood versus adult mortality.

The impact of PopEVE extends beyond just identifying known disease links. In testing, it demonstrated the ability to uncover over a hundred novel disease-causing alterations and new candidate genes previously not associated with developmental disorders. Furthermore, a crucial aspect of PopEVE's design is its commitment to reducing ancestry bias, a common problem in genetic databases that can lead to misdiagnosis in underrepresented populations. By treating all human variants equally, regardless of their prevalence in specific ancestral groups, PopEVE offers a more equitable and powerful tool for genetic interpretation, ultimately paving the way for better treatments and drug discovery for those affected by rare genetic conditions.