The COVID-19 vaccine rollout, while a scientific triumph, revealed the decades-long process of vaccine development can be glacially slow. Artificial intelligence (AI) is now poised to drastically shorten this timeline for future pandemics. Researchers are utilizing AI, particularly machine learning, to rapidly identify vaccine candidates, accelerate preclinical phases, and potentially improve the efficacy of existing vaccines like the flu shot, moving vaccinology into a new, faster era.
Saving money by failing faster
Developing a novel vaccine costs an average of $886.8 million, a price that machine learning aims to reduce by accelerating drug and vaccine discovery. Major pharmaceutical companies like Pfizer, Eli Lilly, and Moderna are investing in AI to streamline this process. While AI-discovered drugs haven't yet passed clinical trials, experts believe AI's current strength lies in helping researchers quickly identify and discard unsuitable candidates, thereby enabling them to 'fail faster' and move more efficiently toward promising lines of inquiry.
AI might outdo scientists at designing flu shots
AI is showing significant promise in improving existing vaccines, such as the annual influenza shot. A 10-year study found that the VaxSeer machine learning algorithm was more accurate than traditional WHO recommendations in predicting dominant flu strains and selecting effective antigenic matches. This enhanced accuracy is attributed to training VaxSeer with vast global influenza surveillance data and information on how antibodies bind to various viral strains, suggesting AI could significantly boost flu vaccine efficacy.
Navigating diverse datasets
A significant challenge in integrating AI into vaccine research is the difficulty in accessing and standardizing diverse datasets, with intellectual property rights often restricting data sharing. Researchers emphasize the need for unified data annotation to prevent algorithmic confusion and flawed results. Even with improved data access, current AI models might still yield multiple promising candidates rather than a definitive few, highlighting that hands-on experimentation remains crucial and will likely experience a bottleneck despite AI's advancements.
Bacterial vaccines are tricky, even with AI
Developing bacterial vaccines presents a greater challenge for AI compared to viral vaccines due to the more complex and varied immune evasion mechanisms of bacterial pathogens. Researchers, like those at the Oxford Vaccine Group working on a Staphylococcus aureus vaccine, believe AI can still help, but it requires significantly larger datasets to recognize intricate patterns. Over time, algorithms are expected to become more sensitive, potentially requiring less data and even predicting immune responses for pathogens, similar to generative AI models.
AI will never completely replace humans
Despite AI's potential, researchers universally agree that human involvement will remain essential in vaccine development, particularly for ensuring safety and efficacy through clinical trials. Concerns about public distrust in AI, similar to that seen with mRNA technology, underscore the need for scientists to effectively communicate with the public. While AI is a powerful tool, it may not fully grasp all biological complexities, maintaining the critical role of human expertise, even if AI eventually streamlines processes to the point of potentially reducing the need for animal trials.
