Proteins are way over nutrients we track on a food label. Present in every cell of our bodies, they work like nature’s molecular machines. They walk, stretch, bend, and flex to do their jobs, pumping blood, fighting disease, constructing tissue, and plenty of other jobs too small for the attention to see. Their power doesn’t come from shape alone, but from how they move.
Lately, artificial intelligence has allowed scientists to design entirely latest protein structures not present in nature tailored for specific functions, corresponding to binding to viruses, or mimicking the mechanical properties of silk for sustainable materials. But designing for structure alone is like constructing a automobile body with none control over how the engine performs. The subtle vibrations, shifts, and mechanical dynamics of a protein are only as critical to its functions as its form.
Now, MIT engineers have taken a serious step toward closing the gap with the event of an AI model generally known as VibeGen. If vibe coding lets programmers describe what they need after which AI generates the software, VibeGen does the identical for living molecules: specify the vibe — the pattern of motion you would like — and the model writes the protein.
The brand new model allows scientists to focus on how a protein flexes, vibrates, and shifts between shapes in response to its environment, opening a brand new frontier within the design of molecular mechanics. VibeGen builds on a series of advances from the Buehler lab in agentic AI for science — systems wherein multiple AI models collaborate autonomously to unravel problems too complex for any single model.
“The essence of life at fundamental molecular levels lies not only in structure, but in movement,” says Markus Buehler, the Jerry McAfee Professor of Engineering within the departments of Civil and Environmental Engineering and Mechanical Engineering. “All the things from protein folding to the deformation of materials under stress follows the basic laws of physics.”
Buehler and his former postdoc, Bo Ni, identified a critical need for what they call physics-aware AI: systems able to reasoning about motion, not only snapshots of molecular structure. “AI must transcend analyzing static forms to understanding how structure and motion are fundamentally intertwined,” Buehler adds.
The brand new approach, described in a paper March 24 within the journal uses generative AI to create proteins with tailor-made dynamics.
Training AI to take into consideration motion
The revolution in AI-driven protein science has been, overwhelmingly, a revolution in structure. Tools like AlphaFold solved the decades-old problem of predicting a protein’s three-dimensional shape. Existing generative models learned to design latest shapes from scratch. But in specializing in the folded snapshot — the protein frozen in place — the sphere largely put aside the property that makes proteins work: their motion. “Structure prediction was such a grand challenge that it absorbed the sphere’s attention,” Buehler says. “But a protein’s shape is only one frame of a for much longer film, and the design space extends through space and time, where structure sits on a wider manifold.” Scientists could design a protein with a selected architecture. They couldn’t yet specify how that protein would move, flex, or vibrate once it was built.
VibeGen does something no protein design tool has done before. It inverts the standard problem. Reasonably than asking, “What shape will this sequence produce?” it asks, “What sequence will make a protein move in precisely this fashion?”
To construct VibeGen, Buehler and Ni turned to a category of AI diffusion models, the identical underlying technology that powers AI image generators capable of making realistic pictures from pure noise. In VibeGen’s case, the model starts with a random sequence of amino acids and refines it, step-by-step, until it converges on a sequence predicted to vibrate and flex in a targeted way.
The system works through two cooperating agents that design and challenge one another. A “designer” proposes candidate sequences geared toward a goal motion profile. A “predictor” evaluates those candidates, asking whether or not they’ll actually move the way in which the designer intended. The 2 models iterate forwards and backwards like an internal dialogue, until the design stabilizes into something that meets the goal. By specifying this vibrational fingerprint because the design input, VibeGen inverts the same old logic: dynamics becomes the blueprint, and structure follows.
“It’s a collaborative system,” Ni says. “The designer proposes, the predictor critiques, and the design improves through that tension.”
Most sequences VibeGen produces are entirely de novo, not borrowed from nature, not a variation on something evolution already made. To substantiate the designs actually work, the team ran detailed physics-based molecular simulations, and the proteins behaved exactly as intended, flexing and vibrating within the patterns VibeGen had targeted.
Considered one of the study’s most striking findings is that many various protein sequences and folds can satisfy the identical vibrational goal — a property the researchers call functional degeneracy. Where evolution converged on one solution, VibeGen reveals a whole family of alternatives: proteins with different structures and sequences that nonetheless move in the identical way. “It suggests that nature explored only a fraction of what’s possible,” Buehler says. “For any given dynamic behavior, there could also be a big, untapped space of viable designs.”
A brand new frontier in molecular engineering
Controlling protein dynamics could have wide-ranging applications. In medicine, proteins that may change shape on cue hold enormous potential. Many therapeutic proteins work by binding to a goal molecule — a virus, a cancer cell, a misfiring receptor. How well they bind often depends not only on their shape, but on how flexibly they will adapt to their goal. A protein that’s engineered with motion could grip more precisely, reduce unintended interactions, and ultimately turn out to be a safer, simpler drug.
In materials science, which is an area of Buehler’s research, mechanical properties on the molecular scale affect their performance. Biological materials like silk and collagen get their strength and resilience from the coordinated motion of their molecular constructing blocks. Designing proteins which are stiffer, flexible, or vibrate in a certain way may lead to latest sustainable fibers, impact-resistant materials, or biodegradable alternatives to petroleum-based plastics.
Buehler envisions further possibilities: structural materials for buildings or vehicles incorporating protein-based components that heal themselves after mechanical stress, or that adjust in response to heavy load.
By enabling researchers to specify motion as a direct design parameter, VibeGen treats proteins less like static shapes and more like programmable mechanical devices. The advance bridges artificial intelligence, medicine, synthetic biology, and materials engineering — toward a future wherein molecular machines could be designed with the identical precision and intentionality as bridges, engines, or microchips.
VibeGen can enterprise into uncharted territory, proposing protein designs beyond the repertoire of evolution, tailored purely to our specifications. It’s as if we’ve invented a brand new creative engine that designs molecular machines on demand,” Buehler adds.
The researchers plan to refine the model further and validate their designs within the lab. In addition they hope to integrate motion-aware design with other AI tools, constructing toward systems that may design proteins to be not only dynamic, but multifunctional; machines that sense their environment, reply to signals, and adapt in real-time.
The word “vibe” comes from vibration, and Buehler sees the connection as greater than wordplay. “We have turned ‘vibe’ right into a metaphor, a sense, something subjective,” he says. “But for a protein, the vibe is the physics. It’s the actual pattern of motion that determines what the molecule can do, the very machinery of life.”
The research was supported bythe U.S. Department of Agriculture, the MIT-IBM Watson AI Lab, and MIT’s Generative AI Initiative.
