Baker Lab Publishes Two Nature Papers on De Novo AI-Designed Proteases

On October 14, the team of David Baker, 2024 Nobel laureate in Chemistry, published two Nature papers in a single day, using generative AI to design from scratch proteases ("molecular scissors") that do not exist in nature, capable of cleaving proteins linked to Alzheimer's disease and ALS.

Confirmed

• The two papers focus on the de novo computational design of metalloproteases and cysteine proteases, respectively
• Metalloproteases: 13 of 69 designs succeeded (roughly 36% success rate); the best design accelerated the reaction about 100 million-fold, and after modifying 4 additional sites reached about 10 billion-fold — though most designs still underperform natural enzymes
• Cysteine proteases: 19% success rate, functional in mammalian cells; the method borrows catalytic atomic geometry from natural enzymes and uses generative models such as RFdiffusion for design, with peptide bonds having a half-life of up to hundreds of years in neutral water — a rigorous test for generative AI
• The metalloprotease paper (Chen et al.) also revealed that a single overlooked water molecule in structure prediction is enough to mislead AI enzyme design — the designed enzyme uses zinc to activate a water molecule to cleave protein chains, and the water must be correctly positioned relative to the peptide; after correcting this, the active-hit rate jumped from 15% to 88%

Why it matters

• This is seen as a landmark step for generative AI in protein design, moving from "folding prediction" to "creating entirely novel functional molecules"
• Some online discussion questioned whether this counts as a "real AI" achievement; @gerardsans and other authors relayed this debate, showing the field remains divided on how to position generative design methods

2026-10-15 ~ 2026-10-15 · 6 related posts

• Baker Lab Publishes Two Nature Papers on De Novo Design of Metallo- and Cysteine Proteases — ArashVahdat · 2026-10-15
• Nobel laureate David Baker uses AI to design novel enzymes cutting Alzheimer's proteins — fanofaliens · 2026-10-15
• [source] Diffusion AI designs working enzymes from scratch: 13 of 69 designs cut peptide bonds — bravo_abad · 2026-10-15
• [source] One water molecule swings AI enzyme design hit rate from 15% to 88%, Nature paper shows — bravo_abad · 2026-10-15
• AI-designed molecular scissors cut Alzheimer's proteins — but is it really AI? — gerardsans · 2026-10-15

1 near-duplicate retellings: IgorCarron