Better expectations: McMaster researcher uses AI to predict pregnancy complications

As a child, Samantha Wilson questioned everything. She was inquisitive, wanted answers and was unhappy when she couldn’t get them.

Now, as an assistant professor in the Department of Obstetrics and Gynecology at McMaster University, Wilson’s career is dedicated to asking questions and using machine learning to find the answers. At the Wilson Pregnancy Lab, she and her team of researchers are working to identify which pregnancies will develop complications with the goal of prevention.

The journey to academia

Wilson’s path to academia and scholarship wasn’t linear. “I had no plans to get here. I wasn’t like, ‘I’m going to be a professor one day,’” she says. “I always tell my students – I didn’t have straight A grades.”

In high school, Wilson became interested in genetics and appreciated the applicability of it – that she could learn a concept and then apply it to other things. After completing an undergraduate degree and travelling, Wilson began a master’s program and was convinced that would be the end of her academic pursuits. That changed when she dug into the research.

“I realized how much I enjoyed it,” she says. Wilson transferred into a PhD in Medical Genetics.

“It just kind of happened where I fell in love with the research and there were questions that I wanted to answer,” she says. “I really like developmental biology and placental biology, not only because I find the topic fascinating, but the entire community surrounding it is just really smart and are nice people to work with who are super helpful and inclusive.”

Cracking the code 

Wilson’s mentors encouraged her to consider a postdoctoral fellowship, focusing on developing machine learning techniques to better understand biology. As a self-taught coder, she remembers worrying she wasn’t qualified, but her mentors and supervisors were confident in her abilities.

“I had never taken a computer science course in my life. I still haven’t!”

Five years later, Wilson leads a lab that primarily focuses on bioinformatics.

“It’s all coding,” she says. The lab uses large, high-dimensional data sets to understand how the placenta develops and functions, and then uses those same data sets and patterns to develop predictive models on pregnancy outcomes.

Predicting complications before they happen

Currently, no test can predict a pregnancy complication before it happens, and Wilson acknowledges the heartbreak and devastation such a test could prevent. Fifteen to 25 per cent of pregnancies end in miscarriage, and yet stigma and shame around pregnancy loss persists.

“The thing I often see in terms of infant and fetal loss is that there’s a tendency for the blame to be placed on the pregnant individual, and so the one thing I would really like to drive home is there is nothing that they did wrong,” she says. “There is nothing that they could have done to prevent it and we’re trying to tackle the underlying molecular structures to understand why this happened, but it’s not their fault.”

Literature shows that changes at a molecular level occur in the placenta when there’s a complication, however studying the placenta early in gestation can be invasive, potentially unnecessary, and risky for the baby. So instead, Wilson and her team study blood and use artificial intelligence (AI) to find abnormalities and patterns that are associated with pregnancy complications.

“Our vision is: at the time where they’re already taking blood in the second trimester for a Maternal Serum Screen, can we just take an extra tube of blood and test it for the risk of preeclampsia, or fetal growth restriction? And the idea is if we have that information, even if there’s no cure for those right now, at least you can increase the monitoring.”

Preventing complications will help optimize maternal and fetal health outcomes, which is important for the individual and their loved ones. But Wilson also points out the economic benefit of this prevention.

“The cost of medical treatment for these pregnancy complications and long-term stays in the neonatal intensive care unit are quite large, and so if we’re able to prevent that from happening, that there’s also an economic benefit to Canada,” she says.

The unpredictable nature of the work is motivating, says Wilson, as is the potential to uncover the mysteries of pregnancy complications and placental dysfunction.

“But the reality is that if I find the answer to that question, I’ll just have a hundred more questions.”