2026年10月21日，西湖大学刘建明等科学家在《自然》发表研究，开发了Pinal模型，实现了从自然语言描述可编程设计功能性蛋白质。

编程生物功能——设计定制蛋白质以执行特定分子任务——是分子工程的一项基础目标。然而，当前的设计模型仍存在根本性局限：它们通常要么需要天然蛋白质作为优化起点，要么需要将功能目标手动重新表述为几何和序列层面的约束来指导候选生成。在此，研究人员介绍了Pinal，一个160亿参数的基础模型，能够根据所需功能的自然语言描述设计候选蛋白质。Pinal在17亿对合成注释的蛋白质-文本对上训练，将功能意图与蛋白质序列和结构联系起来。在计算评估中，生成的候选蛋白质兼具高预测可折叠性、功能描述对齐性和序列多样性，为优先筛选可实验测试的设计提供了基础。研究人员将Pinal应用于四项不同的设计任务——一种荧光蛋白、一种聚对苯二甲酸乙二醇酯水解酶、一种醇脱氢酶和一种代谢H蛋白——并在每种情况下通过实验观察到预期功能，包括两种设计酶的催化活性。在没有迭代实验优化的情况下，一种由Pinal设计的H蛋白在多酶CO?固定途径中，使产物滴度相对于相应的大肠杆菌天然H蛋白提高了1.7倍。这些发现支持自然语言作为蛋白质设计中候选生成的高层界面，使得能够以减少对手动指定结构或序列约束的依赖进行可编程探索。

附：英文原文

Title: Programmable design of functional proteins from natural language

Author: Dai, Fengyuan, You, Shiyang, Zhu, Yudian, Gao, Yuan, Fu, Lihao, Zhou, Xibin, Su, Jin, Wang, Chentong, Fan, Yuliang, Tan, Qingsong, Ma, Xiaoxiao, Deng, Xianjun, Yu, Letong, Qian, Hui, He, Yan, Ke, Yitao, Zhang, Zhe, Han, Chenchen, Chang, Xing, Zheng, Liangzhen, Wang, Sheng, Ji, Yuan, Hao, Kexin, Cao, Longxing, Wang, Yajie, Zeng, Anping, Wang, Shunzhi, Zhan, Chunjun, Zhao, Jing, Si, Tong, Liu, Jianming, Lu, Hongyuan, Yuan, Fajie

Issue&Volume: 2026-10-21

Abstract: Programming biological function—designing bespoke proteins to perform specified molecular tasks—is a foundational goal of molecular engineering. However, current design models remain fundamentally limited1,2,3: they typically require either natural proteins as starting points for optimization or manual reformulation of functional goals as geometric and sequence-level constraints to guide candidate generation. Here we introduce Pinal, a 16-billion-parameter foundation model that designs candidate proteins from natural-language descriptions of desired function. Trained on 1.7 billion synthetically annotated protein–text pairs, Pinal links functional intent to protein sequence and structure. In computational evaluations, generated candidates combined high predicted foldability with functional-description alignment and sequence diversity, providing a basis for prioritizing experimentally testable designs. We applied Pinal to four distinct design tasks—a fluorescent protein, a polyethylene terephthalate hydrolase, an alcohol dehydrogenase and a metabolic H-protein—and experimentally observed the intended function in each case, including catalytic activity for both designed enzymes. Without iterative experimental optimization, a Pinal-designed H-protein increased product titre by 1.7-fold relative to the corresponding native Escherichia coli H-protein in a multi-enzyme CO2 fixation pathway. These findings support natural language as a high-level interface for candidate generation in protein design, enabling programmable exploration with reduced reliance on manually specified structural or sequence constraints.

DOI: 10.1038/s41586-026-11205-z

Source: https://www.nature.com/articles/s41586-026-11205-z