Advances in genome sequencing are giving more families access to prenatal genetic testing and new information about an unborn baby's health, including whether genetic changes may link to a neurodevelopmental condition.
But testing can also uncover DNA changes that clinicians do not yet understand. Known as variants of uncertain significance or VUS, these changes can't be easily classified as harmful or harmless based on available evidence. The result: uncertainty for a significant number of expecting parents.
"Doctors may detect a new genetic change, but in about one-third of cases they won't know if it's truly abnormal or simply within the normal sphere of development," says Dr. Rosanna Weksberg , Clinical Geneticist and Senior Associate Scientist, Genetics & Genome Biology at The Hospital for Sick Children (SickKids). "With too many variants for geneticists to keep up with, it's impossible for them to always know if a variant of interest may result in neurodevelopmental or other genetic disorders."
Members of the Weksberg lab are experts in epigenetics, the study of factors that influence how genes are turned on and off. Their newest innovation is a machine learning model that can help clinicians assess uncertain variants no matter what samples they are able to use and, as a result, help bring long-needed clarity to families.
Transforming the use of 'episignatures'
Episignatures are distinct chemical tags in DNA that reveal the presence of specific genetic conditions. In 2020, the team unveiled a platform, EpigenCentral , that uses blood-derived episignature data to help doctors determine if a variant is disease-causing or benign.
To date, the Weksberg lab has helped establish more than 60 episignatures, most clinically proven for making diagnoses. Until now, these patterns have been tissue-specific, meaning that one found in blood could not be applied to other tissues, leaving this valuable diagnostic approach out of reach when testing amniotic fluid or placental tissue in prenatal settings.
"If we could take these blood-derived signatures and make them tissue-agnostic, we could overcome one of the biggest limitations in prenatal diagnostics," says Dr. Sanaa Choufani, Senior Research Associate.
This inspired their new machine learning method, which could transform blood-derived episignatures into "tissue-agnostic" ones (those that stay the same no matter where they originated in the body). That is, they are informative in multiple tissue types including prenatal samples. Weksberg and Choufani co-led a new study in The American Journal of Human Genetics demonstrating that this approach indeed works, opening the door to its future diagnostic use in prenatal medicine.
New model proves effective in identifying variants
In proving the concept, the team generated a blood-derived episignature using samples from 266 people with Down syndrome, selected because it is one of the more common rare genetic conditions. They then trained their model using publicly available DNA methylation data (chemical tags) from 850 individuals with and without Down syndrome, covering six different prenatal and postnatal tissue types.
Their model accurately recognized the Down syndrome pattern in every tissue type tested, showing that a blood-derived episignature can be changed into one capable of identifying a disease-specific episignature across different tissues.
"We're thrilled that our model can bring a new level of precision to prenatal testing, where so many questions remain to be answered," Weksberg says. "This machine learning approach is also a building block to study many disorders where tissues are inaccessible, which would support rapid translation into the clinic."
In fact, the researchers say this could eventually support diagnostic testing using a broader range of samples, such as saliva and oral swabs — thereby reducing the need for blood or other tissue.
In overcoming a key limitation of epigenetic testing, the team hopes to one day enable earlier, more accurate diagnoses and help families make decisions with clearer information. The work supports Precision Child Health , a movement at SickKids to support individualized care for each child, including better diagnosis and prediction.
