Decades after its discovery, HIV continues to take lives and challenge global health systems. There are many challenges to combating HIV, including the complexity of developing a vaccine for a virus that mutates rapidly, constantly changing its form and evading the immune system. Another barrier is the amount of data that need to be analyzed, from clinical trials and other research, in order to make progress.
Now, Scripps Research scientists are hoping to clear these hurdles through the purchase of high-performance computing equipment. The equipment will be used to accelerate the identification of more effective HIV vaccine candidates through enhanced computational infrastructure, reduced data-processing bottlenecks, and state-of-the-art artificial intelligence (AI) technology. The money needed for the purchase—$1.1 million—comes from the Scripps Consortium for HIV/AIDS Vaccine Development (CHAVD) which is supported by the National Institutes of Health (NIH).
“Over the last 10 years, we’ve been able to accelerate data generation, but we don’t have a good way of analyzing that data to understand if these vaccines are working well,” says Bryan Briney, PhD, associate professor at Scripps Research and co-principal investigator on the project. “This new AI technology will supercharge our ability to evaluate up to millions of potential vaccine designs in the time it used to take to study a few dozen—bringing us closer to finding more promising vaccine approaches.”
The Scripps Research team hopes to eventually develop a long-lasting vaccine that adapts to mutations of the virus and can be delivered in a single dose. In the meantime, however, Briney and the collaborators aim to develop a series of multiple vaccines that adapt to the virus’ changes over time. To meet the challenge of protecting against more than 90% of HIV strains, the team needs real-time feedback from clinical trials—data that reveal how the vaccine is performing and informs the design of the next version in the series.
“We’re shifting from trial-and-error to smart prediction,” says Andrew Ward, PhD, professor in the Department of Integrative Structural and Computational Biology and co-principal investigator on the project. “Instead of spending months testing every design idea in the laboratory, we can screen hundreds of thousands of possibilities computationally, identify the best candidates, and focus our experimental work where it matters most.”
The funds will be used to purchase new AI technology that doubles the computational power available at Scripps Research and operates at speeds four-to-five times faster than existing systems. This new computational bandwidth will allow the team to rapidly analyze the antibodies produced by people who receive experimental vaccines in clinical trials and determine if they are on the right track with molecular precision. The team will evaluate vaccine-induced antibodies, test multiple scenarios simultaneously, and model how they interact with the virus at the molecular level. The antibodies identified to work exceptionally well against the virus will make up the next iteration of the vaccine.
The teams will first train the AI system on historical clinical trial data from previous vaccines to develop a comprehensive computational model that can quickly identify the best antibody candidates. To further develop this AI framework, the group will leverage the StepwiseDesign method which mimics how the immune system gradually learns to develop more efficient antibodies through small, optimized iterations.
The approach has already proven successful: the team used their AI system to analyze about 2,000 antibodies from people who had never been infected with HIV, searching for rare candidates that might have the potential to fight the virus. They discovered an antibody that could neutralize HIV—the first time anyone has found such an antibody in an uninfected person. This finding is significant because it demonstrates that some people naturally carry the genetic starting material for broadly protective antibodies, even though they’ve never encountered HIV.
A successful vaccine would need to activate and train these rare precursor antibodies to mature into full-strength virus fighters. The discovery also validates that this computational approach can identify these extremely rare candidates—essentially finding needles in biological haystacks—which gives scientists confidence the methods will work even better for evaluating antibodies that have already been partially trained by experimental vaccines.
Several HIV vaccine candidates are currently being tested in human trials, producing a flood of new data. With the ability to rapidly analyze these responses and refine follow-up vaccines, researchers could significantly shorten the path to an effective HIV vaccine.
