Data experts from the Icahn School of Medicine at Mount Sinai, New York, have crafted an AI model that can predict which drugs, not yet flagged as dangerous, might cause birth defects. This groundbreaking research, published in the Communications Medicine journal, could be a game-changer in ensuring the safety of medications for pregnant women.
What Exactly Did They Create? This “knowledge graph” model doesn’t just focus on existing medicines. It’s designed to foresee the risks of pre-clinical compounds—those still in the research phase—that could harm an unborn baby. It’s like having a crystal ball that offers a peek into potential medical risks before they become widespread issues.
Why Is This So Important? Approximately 1 in 33 babies in the U.S. is born with a birth defect. These can be functional (how the body works) or structural (how the body is formed). While genetics play a role, the causes for many of these conditions are still a mystery. However, substances in our daily lives—from medications to cosmetics and even foods—can be culprits if pregnant individuals are exposed to them.
Dr. Avi Ma’ayan from Mount Sinai puts it simply: “Our goal is to better predict and protect against the harm that birth defects can cause by understanding and anticipating the effects of new drugs before they hit the market.”
How Did They Do It? The team collected and integrated multiple data sets related to birth defects, genetics of reproductive health, medicine classifications during pregnancy, and how drugs affect our cells. By combining this information, they created a model that can predict if a drug is likely to cross the placenta and potentially cause birth defects.
A Peek Into Their Methods: They used a method called ReproTox-KG combined with semi-supervised learning (a type of AI that uses a small amount of known data to make predictions about larger unknown data). This helped them evaluate over 30,000 preclinical drugs. They also identified patterns linking birth defects, genes, and drugs that might explain the biological reasons some drugs cause birth defects.
A Word of Caution: As exciting as these findings are, the researchers stress they’re preliminary. More research is required to validate them.
What’s Next? The team is planning more projects using this method to delve deeper into the connections between genes, drugs, and diseases. They also want to use this data for educational purposes, teaching others about bioinformatics analysis.
Dr. Ma’ayan envisions a future where regulatory bodies, like the U.S. Food and Drug Administration, might use this model to evaluate the safety of new drugs. He says, “We’re paving the way for a new global framework to assess drug safety and understand why some cause birth defects.”
Final Thoughts: This exciting development is a step towards a safer future for expectant mothers and their babies. By harnessing the power of AI, we’re getting closer to understanding and preventing potential threats before they become widespread problems.