Antibody design across observed and forecasted antigenic evolution
The paper introduces "Germinal Multi," a computational framework that successfully designs and rescues antibodies by jointly optimizing a single paratope against multiple observed and forecasted antigenic variants, thereby generating broad therapeutic efficacy against both current and prospective viral and cancer targets.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Antibodies are the immune system's precision tools, proteins that patrol the bloodstream looking for specific invaders like viruses or cancerous cells. When an antibody finds its target, it latches on with a unique shape, much like a key fitting into a lock, and neutralizes the threat. This mechanism has become a cornerstone of modern medicine, offering treatments for everything from infections to cancer. However, the targets these antibodies seek are not static. Viruses and cancer cells constantly change their outer surfaces through a process of evolution, accumulating small mutations that can alter the shape of the "lock." If the lock changes shape even slightly, the antibody key may no longer fit, rendering the treatment useless. This arms race between our defenses and evolving pathogens is a major reason why some medicines stop working over time.
Scientists have long sought a way to design antibodies that can keep up with these changes, creating keys that fit not just one lock, but many different versions of the same lock. The challenge has been that designing a single antibody to recognize multiple, slightly different shapes is incredibly difficult. Traditional methods often rely on screening vast libraries of natural antibodies or modifying existing ones, a process that can be slow and resource-intensive. Furthermore, predicting how a virus might change in the future to evade our current defenses adds another layer of complexity. Researchers need a way to design antibodies that are robust against both the variations we see today and the potential changes that might emerge tomorrow.
A team of researchers has developed a new computational framework called Germinal Multi to tackle this problem. Instead of designing an antibody for a single target, this system allows scientists to design a single antibody that is optimized to bind to multiple versions of a target simultaneously. The researchers tested this approach in three distinct scenarios: against cancer-related proteins that have developed resistance to existing drugs, against historical versions of a virus that have already changed, and against a predicted future version of a virus that does not yet exist in the wild. In each case, the goal was to create a single antibody that could recognize and neutralize a diverse set of targets, effectively staying one step ahead of evolution.
The first test case involved a protein found on the surface of cancer cells called EGFR. Certain mutations in this protein allow cancer cells to resist treatment with a common antibody drug called cetuximab. The researchers identified six specific mutations known to cause this resistance and asked the computer to design a new antibody that could bind to the original protein as well as all six mutated versions. They did not design the antibody to fit each mutation individually; instead, they created a single design process that considered all of them at once. The computer generated 60 potential antibody sequences, and after testing a small number of the most promising candidates in the lab, the team found several that worked. Four of these new designs bound strongly to the original protein and all six resistant variants, whereas the original drug failed to bind to most of them. This demonstrated that it is possible to engineer a single antibody that ignores the specific changes cancer cells use to hide from treatment.
The second application focused on the SARS-CoV-2 virus, specifically the part of the virus that allows it to enter human cells. The researchers chose two distinct versions of the virus that had appeared at different times: one from an earlier wave of the pandemic and a newer, more evolved version that had already escaped the protection of a known antibody. They used the computer to design a new antibody from scratch that would bind to both versions simultaneously. After screening 57 designs, they identified a candidate that successfully neutralized both virus variants in laboratory tests. Remarkably, this new antibody also worked against several other intermediate versions of the virus that were not part of the original design process, showing that the approach could generate broad protection across a range of evolutionary steps.
The most forward-looking experiment involved predicting how the virus might change in the future. The researchers used a separate computer model to forecast which mutations were likely to appear next in a currently circulating virus lineage. They combined these predictions with data from previous experiments to identify a specific set of changes that would likely allow the virus to escape a powerful, broad-spectrum antibody. They then built a synthetic version of the virus with these predicted changes, creating a "challenge" variant that the existing antibody could no longer neutralize. Using their multi-target design system, they redesigned the original antibody to recognize both the current virus and this future, engineered threat. The result was a new antibody that neutralized the current virus even more effectively than the original and also neutralized the future challenge variant. This proved that evolutionary forecasting could be used to guide the redesign of medicines before the threat even appears in nature.
Throughout these experiments, the researchers found that designing for multiple targets at once was more effective than trying to solve each problem separately. The computer models successfully identified a single antibody shape that could accommodate the differences between the various targets. The team screened tens of designs for each campaign, a number far smaller than the thousands often tested in traditional methods, yet they still found highly effective candidates. This suggests that computational design can efficiently navigate the complex space of possible antibody shapes to find those with the desired breadth of activity. The work also highlighted that binding to a target does not always guarantee that the antibody will neutralize the virus, as seen when one candidate bound to the virus but failed to stop it from infecting cells. This underscores the importance of testing the final designs in functional assays that measure actual neutralization.
The study concludes that it is possible to move beyond designing antibodies for a single snapshot of a virus or cancer cell. By treating the design process as a multi-objective problem, scientists can create therapeutics that are resilient to the natural changes of their targets. The ability to incorporate predictions about future evolution into the design process offers a new path for proactive medicine, potentially allowing us to prepare treatments for threats that have not yet emerged. While these results are from laboratory studies and require further validation for clinical use, they establish a powerful new strategy for generating antibodies that can withstand the relentless pressure of evolution.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.