How Scientists Design Proteins From Scratch

The Building Blocks of Life, Reimagined

Every living cell runs on proteins. These intricate molecular machines digest food, fight infections, carry oxygen, and power every contraction of the heart. For billions of years, evolution alone decided what proteins existed. Now, for the first time, scientists are writing their own.

The field is called de novo protein design — from the Latin for "from new" — and it allows researchers to create proteins that have never existed in nature, engineered atom by atom on a computer and then brought to life in a laboratory. The achievement was recognized in 2024 when biochemist David Baker of the University of Washington shared the Nobel Prize in Chemistry for his pioneering work in this area.

What Exactly Is a Protein?

Proteins are long chains of amino acids — small chemical building blocks — that fold into precise three-dimensional shapes. That shape determines everything: how a protein behaves, what it binds to, and what it does. Hemoglobin curves into a pocket that grips oxygen molecules. Antibodies form Y-shaped clamps that latch onto viruses. Enzymes create active sites that accelerate chemical reactions with extraordinary precision.

Nature selects protein sequences through billions of years of trial-and-error evolution. De novo design flips this process: scientists start with the shape they want and work backwards to find an amino acid sequence that will fold into it.

How De Novo Protein Design Works

The process combines computational modeling, physics, and machine learning. According to the University of Washington's Institute for Protein Design, the core steps are:

• Define the target function — What should the protein do? Bind a virus? Catalyze a reaction? Fit into a receptor?
• Design the structure — Use software to model a three-dimensional shape capable of performing that function.
• Compute the sequence — Determine which combination of amino acids will reliably fold into that shape.
• Build and test — Synthesize the gene, insert it into a cell, let it produce the protein, and verify it works as predicted.

The first major proof came in 2003, when Baker's team used their Rosetta software to design a small protein called Top7 — the first fully artificial protein ever built. Its structure had no equivalent anywhere in the natural world, yet it folded exactly as predicted.

The AI Revolution: From Rosetta to RFdiffusion

For years, progress was painstaking. Designing even a small functional protein could take months. That changed dramatically with artificial intelligence.

The 2020 release of AlphaFold2 by DeepMind — which also shared the 2024 Nobel — cracked the "protein folding problem," predicting the structure of almost any known protein with near-perfect accuracy. Baker's lab then developed RFdiffusion, a generative AI model that works in reverse: instead of predicting a structure from a sequence, it generates entirely new structures on demand. Published in Nature in 2023, RFdiffusion achieved success rates roughly 100 times higher than earlier methods when designing proteins that bind specific molecular targets.

The tool works similarly to image-generating AI: it starts with random noise and iteratively refines it into a coherent protein backbone, guided by the user's specifications.

What Designed Proteins Can Do

The applications span medicine, industry, and the environment. According to the National Institutes of Health, designed proteins could:

• Fight viruses — Baker's team created a mini-protein of just 56 amino acids that inhibits SARS-CoV-2, the virus behind COVID-19.
• Improve vaccines — Novel proteins can present antigens to the immune system more effectively than natural structures.
• Treat disease — Custom-designed binders can block cancer receptors or deliver drugs with pinpoint precision.
• Clean up pollution — Engineered enzymes can break down plastics, pesticides, and toxic industrial chemicals.
• Enable new materials — Protein-based nanomaterials could transform electronics, sensors, and construction.

In 2026, researchers reported a method for designing protein assemblies regulated by small-molecule drugs — including one controlled by the FDA-approved compound amantadine — opening the door to "programmable" proteins that switch on or off in response to medication.

Why It Matters

Evolution is constrained by what already exists. De novo protein design is not. By building proteins from first principles, scientists can solve problems that nature never encountered — and do it in months rather than millions of years.

As Baker told the Nobel committee, the ability to design proteins at will represents "a new era of molecular engineering." The first fully designed proteins are already entering clinical trials, and the pace of discovery is accelerating. What was once the exclusive domain of evolution now belongs, at least in part, to human ingenuity.