Staying ahead of viruses can be challenging. Can AI help?
Noor Youssef, Scientific lead of the Predictive Modeling for Vaccine Design group at Harvard Medical School, details why the answer may be yes.
Dr. Noor Youssef is a mathematician turned biologist driven by a passion for using mathematics and computer science to tackle pressing challenges in human health. She received her PhD in computational biology from Dalhousie University, where she focused on mechanistic models of evolution, and is currently a Scientific Lead in the Marks Lab at Harvard Medical School. Her work centers on viral evolution and the development of AI-driven tools to guide vaccine and therapeutic design, with a broader goal of building predictive frameworks that can inform real-world health decisions.
Outpacing the Virus How AI Can Future-Proof Vaccines
One big problem with how we currently test vaccines is that we’re always chasing the past. When a new vaccine is developed, we evaluate how well it protects against previous and current versions of a virus—but by the time the vaccine is approved and rolled out, often months later, the virus has already evolved into something new. This delay can lead to a mismatch between the version of the virus in the vaccine and the version that actually ends up circulating. For flu, and now SARS-CoV-2, this mismatch often leads to vaccines that are less effective than they could be.
Our team set out to improve this process by creating an AI model called EVE-Vax that predicts viral evolution and can generate new variations of viral proteins that mimic what future versions of a virus might look like, giving us a hint into what the future might be.
We used EVE-Vax to predict the evolution of SARS-CoV-2 at 5 different time points during the COVID-19 pandemic, and we got really exciting results. Our model designed proteins that closely matched what we ended up seeing in the real world, in terms of how easily the proteins infected human cells, and how well they avoided the immune system.
What’s powerful about EVE-Vax is that it does not require patient samples or costly and time-consuming experiments, which means that it can be applied early before a viral outbreak even begins. We think our model could be a valuable tool to help scientists develop future-proof vaccines for SARS-CoV-2, and for other quickly evolving viruses with pandemic potential. Our hope is that by using our approach, we can stop playing catch-up with viruses by having insights into their potential next move.
