De novo protein design: specifying biology before finding it
How the design-build-test loop works, why structural verification inside the loop keeps the discipline honest, and what the Institute for Protein Design's budget scale and Tianjin University's synthesis throughput establish about the field.
The molecular age searched nature for proteins that did useful things and improved what it found. De novo protein design reverses the direction: the function is specified first — bind this target, catalyze that reaction, self-assemble into that cage — and the sequence is computed from physical principles, yielding an amino-acid string that has never existed in any organism. The claim underneath the discipline is auditable: the mapping from sequence to folded structure is computable, so a design can be falsified. That falsifiability is the whole culture of the field — a designed protein is either verified against its model by cryo-EM or X-ray crystallography, or it is back to the drawing board.
What has changed in the last decade is that the loop became an institution rather than a laboratory feat. The Institute for Protein Design at the University of Washington, directed by David Baker and housed in the School of Medicine, runs the design-build-test cycle at scale: more than 250 members, a $33 million annual research budget plus a $7 million Washington Research Foundation grant in 2026, core laboratories that put protein production, peptide synthesis, electron and light microscopy and X-ray crystallography in one facility stack, fifty-plus publications a year and more than 100 issued patents. The verification step living inside the same building as the design step is not a convenience — it is the reason the loop shortens. Every failed fold feeds the next design generation, which is how single successes become designable families.
The build arm has industrialized on the other side of the Pacific, and its center is Tianjin University: a School of Synthetic Biology and Biomanufacturing housing the State Key Laboratory of Synthetic Biology, the Frontier Science Center for Synthetic Biology and the Key Laboratory of Systems Bioengineering — a national-facility stack dedicated to synthesis, assembly and characterization. Its research runs from yeast genome synthesis to DNA information storage and artificial cell construction, it ranks first globally in Scopus synthetic-biology publications, and it trained roughly 2,000 postgraduates as of 2022 — the workforce a design discipline scales through.
The honest boundary is the distance between a verified structure and a product. A designed therapeutic must still run the biologics gauntlet — FDA review in the US, the ATMP framework of Regulation (EC) No 1394/2007 in Europe, NMPA licensing in China — and the field’s best-funded companies, Generate Biomedicines and Arzeda among them, exist precisely to carry verified designs across that distance, though their evidence base on this site remains ledger-thin. The near-term competition is therefore not who can compute the most backbones but who owns the unglamorous middle: expression yields, stability under formulation, immunogenicity — the product properties no design model scores yet, and the reason the loop still ends in a manufacturing plant rather than a paper.
Evidence and practice
De novo protein design — the technology dossier: sources with check dates, maturity, what it replaces, how to implement it, who delivers it.