General Multimodal Protein Design Enables DNA-Encoding of Chemistry

2026-04-06

AI Summary

Current enzyme design approaches rely on predefined catalytic residues, limiting the expansion of DNA-encodable chemical reaction space. This work proposes DISCO, a multimodal diffusion model that, for the first time, enables joint generation of protein sequences and three-dimensional structures solely from reaction intermediates—without requiring pre-specified catalytic residues—thereby designing multifunctional heme enzymes with novel active-site geometries. The method supports cross-modal, multi-objective optimization during inference and integrates directed evolution for experimental validation. Designed enzymes efficiently catalyze diverse non-natural carbene transfer reactions, including cyclopropanation, B–H insertion, and C(sp³)–H insertion, outperforming existing engineered enzymes in activity, with further enhancements achievable through random mutagenesis.

Abstract

Evolution is an extraordinary engine for enzymatic diversity, yet the chemistry it has explored remains a narrow slice of what DNA can encode. Deep generative models can design new proteins that bind ligands, but none have created enzymes without pre-specifying catalytic residues. We introduce DISCO (DIffusion for Sequence-structure CO-design), a multimodal model that co-designs protein sequence and 3D structure around arbitrary biomolecules, as well as inference-time scaling methods that optimize objectives across both modalities. Conditioned solely on reactive intermediates, DISCO designs diverse heme enzymes with novel active-site geometries. These enzymes catalyze new-to-nature carbene-transfer reactions, including alkene cyclopropanation, spirocyclopropanation, B-H, and C(sp$^3$)-H insertions, with high activities exceeding those of engineered enzymes. Random mutagenesis of a selected design further confirmed that enzyme activity can be improved through directed evolution. By providing a scalable route to evolvable enzymes, DISCO broadens the potential scope of genetically encodable transformations. Code is available at https://github.com/DISCO-design/DISCO.

Problem

• protein design
• enzyme design
• DNA-encoded chemistry
• carbene-transfer reactions
• de novo enzyme

Innovation

• multimodal protein design
• diffusion model
• enzyme design
• carbene-transfer reactions
• sequence-structure co-design