Discovery

Anthropic's Claude autonomously designs protein binders against 14 of 15 targets

Anthropic's Claude autonomously designs protein binders against 14 of 15 targets

A large language model operating without human input into any individual design decision produced confirmed protein binders against 14 of 15 testable targets, including several considered difficult by conventional computational methods, according to a technical report published August 18, 2026 by Anthropic.

The work, led by Amir Shanehsazzadeh at San Francisco-based Anthropic, describes a system in which Claude Opus 4.8 and a second model, Mythos Preview, each ran complete protein binder design campaigns — from target research and epitope selection through structure generation, sequence optimization, and candidate ranking — guided solely by a ~16,000-word protocol prompt. Once the expert-written protocol was fixed, no specialist scientific input was provided during individual design decisions. The authors describe the work as foundational research with no current clinical translation announced.

What the system did

The agent orchestrated a pipeline of open-source tools to generate novel single-chain miniproteins of 50–120 residues designed to bind biologically relevant surfaces on target proteins. Across 16 structurally and functionally diverse targets — including cytokines, cell-surface receptors, viral glycoproteins, and enzymes — the system generated 1,440 designs in total, of which 1,320 were analyzed. Experimental validation was performed independently and blinded by two contract research organizations: Switzerland-based Adaptyv Bio, using cell-free expression and surface plasmon resonance (SPR)/bio-layer interferometry (BLI), and South San Francisco-based Twist Bioscience (Nasdaq: TWST), expressing designs as human IgG1 Fc fusions in HEK293 cells with high-throughput SPR arrays.

Binding was confirmed for designs against 14 of 15 targets with interpretable data; one target, mature GDF-8, was excluded due to aggregation artifacts. The overall hit rate across 1,320 analyzed designs was 27% (354 confirmed binders), rising to 49% for top-ranked designs. Of the 354 confirmed binders, 194 bound below 100 nM, 90 below 10 nM, and 42 below 1 nM. Cross-species reactivity was achieved without explicit per-target engineering: 154 of 179 binders tested also bound the cynomolgus ortholog, and 130 of 233 bound the mouse ortholog.

Difficult targets and a competitive benchmark

Two results stand out scientifically. Against TNF-alpha (TNFα), a compact homotrimer whose receptor-binding grooves span subunit interfaces, the system produced 12 confirmed binders on four distinct backbones, with the tightest apparent KD of 0.70 nM. The authors note that multiple prior de novo design efforts had reported no binders against this surface.

Against RBX1, an E3 ligase subunit that had been the subject of an open community design competition in which 9 of 245 submitted designs bound, Claude's campaigns produced 28 of 90 binding designs, with a top binder at 3.9 nM — compared to 45 nM for the competition's winning entry measured on the same assay plate. The authors note that Claude's designs largely avoided the CUL1-occluded face of RBX1, with only 17% of footprint residues overlapping that region.

The AllSci BriefSystematic R&D and deal news. Daily.

A key technical contribution is the ranking score used to select candidates before synthesis: an ensemble co-folding confidence metric combining interface predicted Structural Alignment Error (ipSAEmin) across three predictors. Validated on a public benchmark of 3,532 designs across 13 targets, this ensemble achieved a macro average precision of 0.66, compared to 0.63 for AlphaFold3 alone on the same benchmark, according to the paper.

Competitive context

Automated protein design is also being advanced through purpose-built approaches including BindCraft, published in Nature in 2025 by the Baker Lab, and Google DeepMind's AlphaProteo. Unlike those systems, Claude acts as an autonomous agent orchestrating specialist protein-design and structure-prediction tools rather than functioning as a purpose-built protein-design model itself. No LLM-agent-designed protein therapeutic has entered clinical trials as of August 2026.

Limitations

The paper's authors acknowledge that the study's evidence is binding, not structure or function. No crystal structures are reported, and no functional inhibition or agonism data are presented. The paper describes a platform, not a drug development program, and no IND filing or regulatory designation has been reported for any binder produced by this system.

The authors note that the identical frozen protocol served all 16 targets, and say this suggests campaigns of this kind should be within reach of laboratories with targets of interest but no expertise in computational protein design. However, the campaigns still depended on an expert-authored protocol, specialized computational tools, and substantial GPU resources.


Spot something wrong? Report an issue with this article