home 回首頁 navigate_next 最新消息 navigate_next 研究成果 navigate_next 本所吳昆峯老師於JACS發表新文章: 人工智慧如何優化蛋白質穩定度的奧義
2026/02/09

本所吳昆峯老師於JACS發表新文章: 人工智慧如何優化蛋白質穩定度的奧義

AI Gets Water Right: A Hydration Shield Forms, Wrapping Proteins in a Protective Coat for a Major Stability Boost

 

A study published in the Journal of the American Chemical Society (JACS) reports that artificial intelligence can enhance protein stability in an unexpected way—by engineering the water around a protein, not just the protein itself. Researchers led by Dr. Kuen-Phon Wu at the Institute of Biological Chemistry, found that AI-designed ubiquitin-fold proteins can achieve exceptional resilience by creating a protective, “mesostructured” hydration shell on their surface.

 

Protein stability under harsh conditions is central to modern bioengineering, enabling more robust therapeutics and industrial enzymes. For decades, the dominant strategy has been to strengthen a protein’s hydrophobic core. But when Wu’s team used the deep-learning design tool ProteinMPNN to redesign ubiquitin (Ub) and related Ub-fold proteins (including ISG15), the resulting variants took a different route. The redesigned proteins—R4, R10, and ICV variants—showed striking resistance to conditions that typically destabilize natural proteins. In stress tests, the AI-generated variants remained folded and functional under extreme heat (reportedly above 120 °C) and under strongly denaturing chemical conditions (a combination of pH 3 and 8 M urea). To uncover the mechanism behind this resilience, the team combined advanced NMR spectroscopy with molecular dynamics simulations. Their analyses indicate that the AI strategically redistributed and clustered surface charges to organize surrounding water into a highly ordered network—a “mesostructured hydration shell.” This structured water layer acts like a hydration shield, helping buffer the protein from thermal and chemical stress and reducing pathways that initiate unfolding.

 

“Some people think ‘Water Breathing’ belongs to fantasy,” the team notes, “but what we’re seeing is designable physical chemistry: by tuning a protein’s surface, AI can make water form a more ordered hydration layer that measurably strengthens stability under extreme conditions.” The findings establish mesostructured hydration as a sequence-encoded, engineerable stability mechanism, opening a new direction for protein design. Beyond reinforcing the “dry” core, future biopharmaceuticals and biocatalysts may be made more durable by commanding the ‘wet’ exterior—the water that surrounds and protects the folded structure.

 

About the Study 

The paper, “Mesostructured Water Enhances Stability of ProteinMPNN-Designed Ubiquitin-Fold Proteins,” was authored by Lu-Yi Chen, Wei-Lin Lu, and colleagues (Academia Sinica, UC Riverside, Osaka University). The work was supported by Academia Sinica Career Development Award (AS-CDA-110-L03), Grand Challenge Seed Grant (AS-GCS-113-L05), Innovative AI Applications in Humanities and Scientific Research (I-AI-A) Projects (AS-IAIA-114-L02), and National Science and Technology Council (NSTC 114-2113-M-001-019). The first-author Lu-Yi Chen was supported by the NSTC Graduate Research PhD Fellowship. This is an open access article (CC-BY-4.0)

 

AI把水用對了:水合護盾成形,蛋白質披上水合外衣、穩定性大幅升級

 

一項已刊登《美國化學會誌》(Journal of the American Chemical Society, JACS)的研究指出,人工智慧(AI)能以出人意料的方式提升蛋白質穩定性——關鍵不只在「把蛋白質本身設計得更緊」,而是把蛋白質周圍的水「設計好」。由生物化學研究所副研究員吳昆峯博士領導的跨國研究團隊發現,AI 設計的 ubiquitin 折疊(ubiquitin-fold)蛋白,能在蛋白質表面形成一層具保護性的「結構化水合外殼(mesostructured hydration shell)」,使其在極端環境下展現卓越耐受性。

 

蛋白質在嚴苛條件下的穩定性,是生物工程的重要核心,直接影響藥物蛋白與工業酵素的可製造性與可保存性。長期以來,主流策略多聚焦於強化蛋白質疏水核心的緊密堆疊;然而,當吳博士團隊使用深度學習蛋白設計工具 ProteinMPNN 重新設計泛素(ubiquitin, Ub)及相關類泛素(包含 ISG15)時,AI 走出了一條不同的路徑。研究團隊得到的設計變體——R4、R10 與 ICV 系列——在多項壓力測試中呈現顯著耐受性。在極端熱條件下(報導顯示可高於 120 °C)以及強去摺疊化學環境中(包含強酸與高濃度尿素),這些 AI 產生的變體仍能維持折疊狀態與功能表現。為釐清其分子機制,團隊結合先進核磁共振(NMR)光譜與分子動力學(MD)模擬解析穩定性來源。結果顯示,AI 透過在蛋白質表面重新分布並聚集電荷,促使周圍水分子形成更有序且緊密的網絡,也就是所謂的「結構化水合層」。這層結構化水分子如同「水合護盾」,可在熱與化學壓力下提供緩衝,降低啟動錯誤摺疊的機會,從而提升整體穩定性。

 

「有人把『水之呼吸』當成想像力,但我們看到的是可被設計的物理化學:AI 透過調控蛋白質表面特徵,讓水合層更有序,進而在極端條件下顯著提升穩定性。」本研究確立「結構化水合層」是一種可被工程化的穩定性機制,為蛋白設計開啟新方向:未來的生物藥物與生物觸媒,除了加固「乾的核心」,也能藉由掌控蛋白「濕的外層」——亦即周圍水分子的組織方式——在製程與儲存的嚴苛環境中更可靠、更耐用。

 

關於本研究

論文題目為 「Mesostructured Water Enhances Stability of ProteinMPNN-Designed Ubiquitin-Fold Proteins」。合作單位涵蓋中央研究院、加州大學河濱分校(UC Riverside)、大阪大學等。研究經費來自 中央研究院前瞻計畫 (AS-CDA-110-L03)、關鍵突破種子計畫(AS-GCS-113-L05)、AI在人文與科學研究的創新應用(AS-IAIA-114-L02)、以及國科會(NSTC 114-2113-M-001-019)。第一作者陳履頤為生化所博士生,獲三年國科會博士生研究獎學金支持。本文章採開放取用出版模式(CC-BY-4.0)

 

 

 

文章資訊 :
期刊: Journal of the American Chemical Society
發表日: 線上ASAP 二月五日
題目: Mesostructured Water Enhances Stability of ProteinMPNN-Designed Ubiquitin-Fold Proteins