Harvard scientists design 'evolutionary traps' to steer viruses away from dangerous variants

For decades, virologists have watched viruses mutate and adapt in real time, always one step behind. A new flu strain slips past the vaccine. A coronavirus subvariant erodes immunity. The world reacts. Harvard researchers have now proposed flipping that dynamic entirely—not by predicting which mutations will emerge, but by reshaping the evolutionary terrain itself so that dangerous variants become evolutionary dead ends.

The method, called fitness landscape design, starts with a metaphor biologists have used for generations. Imagine each viral sequence as a point on a landscape, where height represents how well that virus spreads. Evolution is a population climbing uphill, seeking higher fitness. For most of scientific history, researchers have treated that landscape as fixed, asking only whether a particular mutation would help or harm a virus's ability to replicate. Eugene Shakhnovich, a chemistry professor at Harvard, and Vaibhav Mohanty, an M.D./Ph.D. student, inverted the question: What if we could reshape the landscape itself by choosing the right combination of antibodies?

The team built a mathematical model that connects molecular details to viral fitness. It tracks how a virus's surface protein binds to human cell receptors and to antibodies, then uses those binding strengths to calculate infection rates and growth. They tested the model first on murine norovirus, a common laboratory pathogen that had been evolved in flasks with and without neutralizing antibodies. The model's predictions matched the actual rise and fall of viral strains in those experiments. They then applied the same framework to SARS-CoV-2, combining measurements of how spike variants bind to the human ACE2 receptor with epidemiological data on each variant's real-world success.

With the model calibrated, the researchers tackled the core problem: the inverse. Rather than asking how a virus would evolve on a given landscape, they specified a desirable landscape—one where an entire network of related SARS-CoV-2 variants had suppressed fitness—and then searched for antibody combinations that would produce it. In simulations, certain antibody cocktails could trap viral evolution, converting pathways that normally led to more fit escape variants into routes ending in low-fitness valleys. The approach works like a chess engine calculating several moves ahead, identifying the best antibodies to suppress fitness gains before mutations even appear.

Mohanty described the shift in thinking plainly: instead of reacting to whatever mutation emerges next, the goal is to proactively steer the virus so that its easiest escape routes become disadvantageous. This addresses a real frustration in public health. The World Health Organization updates flu vaccines twice yearly, yet surprise strains still emerge. SARS-CoV-2 has produced a steady stream of subvariants that gradually chip away at vaccine protection. Viruses are constantly mutating to evade immune systems, and vaccines end up playing perpetual catch-up.

The researchers see applications far beyond infectious disease. In cancer immunotherapy, treatments like CAR-T cell therapy arm immune cells with engineered antibody-like receptors to attack tumors that themselves evolve. Fitness landscape design could help design those immune defenses so cancers have fewer escape routes. In industrial biotechnology, scientists use directed evolution to improve enzymes and proteins for medicines and manufacturing, but the trial-and-error approach often gets stuck—finding a protein better than the starting version but far from optimal. The team believes fitness landscape design could guide evolution around these dead ends toward proteins with more useful properties.

The researchers are also incorporating artificial intelligence, including protein language models and generative tools, to predict how antibodies will bind to viral proteins and to search through vastly more possible antibody designs. Shakhnovich noted that the fitness landscape concept itself is old, but what is new is the ability to manually mold that landscape to fight disease. If the method holds up in experimental validation, it could give drugmakers and public health officials a powerful tool in the ongoing viral arms race—one that shifts the advantage from reaction to anticipation.