AI-powered de novo architecturing of marine-mimetic antioxidant peptides in yeast synbio factories

作者：Le Gao, Tiantian Zhou, Jie Yuan, Shouchang Zhang, Kai Hong, Zhaokun Zhang, Lee Chia Hau, Wai San Cheang, Chi-Chun Zhou, Xin Wu · 发表于：Chemical Engineering Journal · 年份：2025 · DOI：10.1016/j.cej.2025.169604 · 被引用次数：12 · 研究领域：Protein Hydrolysis and Bioactive Peptides、Computational Drug Discovery Methods、Neurobiology and Insect Physiology Research

The sustainable biosynthesis of structurally stable bioactive peptides remains a critical challenge for methanol-driven microbial cell factories. Here we present an AI-guided platform enabling de novo design of synthetic oyster-mimetic peptides (SOPs) through evolutionary-inspired protein assembly. A transformer-based language model prioritized 28 antioxidant peptides from 160 candidates, which were computationally assembled into template-free fusion architectures via natural sequence linguistics. SOPs successfully produced in Pichia pastoris under industrial scales, demonstrated a high yield of 34 g/kg dry cell weight, a reduced glutathione (GSH) content of 126.07 mg/L, and notable stability with a half-life of 15 days. Comparison with natural counterparts, SOPs exhibited 703 % higher radical scavenging capacity and superior hepatoprotection function in murine models. The embedded peptide VISDEHGIDPTK in SOPs acts as a Peroxisome proliferator-activated receptor alpha (PPARα) agonist, stabilizing ligand-receptor interactions via allosteric hydrogen bonding with Lys292 and Glu462, thereby ameliorating lipid dysregulation. SOPs demonstrated 90 % lower carbon emissions, 82 % water reduction, and 87.5 % energy savings in life cycle analysis compared to conventional marine extracted natural oyster-mimetic peptides (NOPs). This machine learning framework establishes a carbon-negative paradigm for precision bioengineering of therapeutic marine-mimetic peptides, overcoming historical...