By Erin Jester
A University of Florida Emerging Pathogens Institute and College of Public Health and Health Professions team led a workshop to train international researchers in the use of artificial intelligence to predict, prepare for and respond to infectious disease outbreaks.
Michael Sy, an epidemiology Ph.D. student, along with his counterpart Simone Rancati, a biomedical engineering Ph.D. student at the University of Pavia in Italy, showcased and held a practice session on novel AI methods for molecular epidemiology during the Genomic Data Intelligence workshop in Taiwan in November.
Marco Salemi, Ph.D., interim director of the Emerging Pathogens Institute and professor of experimental pathology in the College of Medicine, was lead organizer of the workshop, co-organizing with the National Taiwan Ocean University in Taiwan, and with the collaboration of faculty from the University of Pavia and UF, including Mattia Prosperi, Ph.D., PHHP’s associate dean for artificial intelligence and innovation, and Simone Marini, Ph.D., an assistant professor in the Department of Epidemiology.
The three-day interdisciplinary workshop covered the principles of molecular epidemiology and AI applications for studying genomic data, with lectures from AI and epidemiology experts from all three institutions.
The GEDI workshop originated through conversations among Salemi and his colleagues at the National Taiwan Ocean University about creating a scholarly exchange program and getting students interested in machine learning in biomedical research.
“The idea is to give students an overview of what we can do in terms of generating data from pathogens,” said Salemi, who is a graduate of the University of Pavia and has continued to collaborate with faculty at the institution.
Salemi invited Sy and Rancati to participate in the workshop based on their previous work with AI-based disease monitoring projects ViraLingo and PhiCoV, headed by Marini and Salemi, respectively. These projects used large language models to analyze viral genomic data and electronic health records to predict which viral strains will be dominant in the future. Sy and Rancati were instrumental in developing the predictive algorithms.
Instead of a reactionary response, where disease monitoring and vaccine development are based on the strain of a pathogen already circulating in the population, this technology could allow researchers to get ahead of the curve and potentially develop vaccines for future strains.
“It gives us a strong advantage,” Salemi said. “Having a tool that can predict which strain is important to select for future vaccines would potentially make it possible to avoid new outbreaks.”
Sy and Rancati presented their work on the AI models with a focus on how to apply them to bioinformatics. The pair also hosted a technical workshop for graduate students describing how to prepare data, how to select a large language model based on the goal of a project, how to train the model and how to interpret results.
Rancati said the workshop was useful to him and to Sy, as they were able to practice explaining the complex subject of AI to participants with purely biological backgrounds. For their part, participants were able to take home knowledge of techniques to improve their research and someday possibly create tools to analyze the spread of viruses.
The workshop was supported in part by a grant from the Global Virus Network, of which the EPI is a part. With dozens of affiliate institutions across the globe, Salemi hopes GVN’s support will allow his team’s methods to reach a broader audience and make the workshop a success.
Sy said getting the information about these AI models out into the world is exactly what he hopes to do.
“AI is here, it’s easy to use, it’s beneficial for research,” he said. “We don’t want to gatekeep this information.”