Researchers at The Hospital for Sick Children have developed a new machine learning model to improve the accuracy of prenatal genetic testing. Currently, clinicians often encounter genetic variants of uncertain significance during testing, which leaves parents without clear answers about their baby's health.
The Weksberg lab at SickKids is using epigenetics to address this problem. Epigenetics focuses on how genes are activated or deactivated, and the team previously created a platform called EpigenCentral to identify disease-causing variants using blood samples. Until now, these chemical signatures were tied to specific tissue types, which limited their use in prenatal diagnostics when doctors are analyzing amniotic fluid or placental tissue.
This new machine learning approach successfully converts blood-derived episignatures into tissue-agnostic markers. By training their model on DNA methylation data from various tissue types, the team proved that these signatures can reliably identify genetic conditions across different parts of the body. In tests, the model accurately recognized patterns associated with Down syndrome in every tissue sample analyzed.
This development is a significant step toward making genetic testing more precise for rare disorders. By expanding the range of samples that can be used for diagnostics, including potentially saliva or oral swabs, this method could reduce the need for more invasive testing procedures. The team aims to help families avoid the prolonged diagnostic uncertainty that often accompanies prenatal findings by providing clearer and more actionable information.