The Chan Zuckerberg Biohub has released the ESM Atlas, an expansive AI-generated database of protein structures.
Chan Zuckerberg Biohub 已發布 ESM Atlas，這是一個由 AI 生成的大型蛋白質結構數據庫。
Introduction
Researchers at the Chan Zuckerberg Biohub have introduced a comprehensive atlas of predicted protein structures and sequences developed via an artificial intelligence model.
Chan Zuckerberg Biohub 的研究人員推出了一份詳盡的預測蛋白質結構與序列圖譜，該圖譜是透過人工智慧模型開發而成的。
Main Body
The ESM Atlas comprises 1.1 billion predicted protein structures and 6.8 billion sequences, significantly exceeding the volume of the AlphaFold Database. This expansion is primarily attributed to the inclusion of metagenomic sequences derived from diverse environmental sources, such as oceanic and soil samples, which were previously underrepresented in existing datasets. The underlying architecture, ESMFold2, utilizes a protein language model trained on evolutionary data. According to Biohub leadership, this system demonstrates superior efficacy compared to AlphaFold3, particularly in the determination of complex protein interactions, including antibody-antigen binding.
ESM Atlas 包含 11 億個預測蛋白質結構與 68 億個序列，規模遠超 AlphaFold 資料庫。此擴展主要歸因於納入了來自多種環境來源（如海洋和土壤樣本）的後基因組序列，而這些序列在先前的資料集中代表性不足。其底層架構 ESMFold2 採用了基於演化數據訓練的蛋白質語言模型。根據 Biohub 領導層的說法，該系統比 AlphaFold3 展現出更高的效能，尤其是在判定複雜蛋白質交互作用（包括抗原抗體結合）方面。
Institutional application of this technology has already manifested in the design of novel protein binders targeting oncological and immunological conditions. Laboratory validation indicates that a substantial proportion of these AI-generated designs function as predicted, specifically in the reactivation of immune cells. Furthermore, the model has facilitated the identification of structural homologies between CRISPR microbial defense proteins and eukaryotic gene-editing proteins. To ensure widespread scientific adoption, the Biohub has opted for an open-source distribution model, providing access via biohub.ai and third-party platforms such as AWS Bio Discovery and SandboxAQ, supplemented by the provision of compute credits for researchers.
該技術在機構端的應用已體現在設計針對腫瘤與免疫疾病的新型蛋白質結合劑上。實驗室驗證顯示，很大比例的 AI 生成設計運作符合預期，特別是在重新激活免疫細胞方面。此外，該模型還促進了 CRISPR 微生物防禦蛋白質與真核生物基因編輯蛋白質之間結構同源性的識別。為確保科學界的廣泛採用，Biohub 選擇了開源分發模式，透過 biohub.ai 及 AWS Bio Discovery 和 SandboxAQ 等第三方平台提供訪問，並為研究人員提供運算額度。
Conclusion
The ESM Atlas is now available as an open-source resource intended to accelerate the discovery of novel biological structures and therapeutic agents.
ESM Atlas 目前已作為開源資源提供，旨在加速發現新型生物結構與治療藥劑。
Vocabulary Learning
The C2 Pivot: Nominalization and the 'Erasure' of Agency
To move from B2/C1 to C2, a writer must transition from describing actions to constructing conceptual frameworks. The provided text is a masterclass in Nominalization—the process of turning verbs (actions) into nouns (entities) to create an aura of objective, academic inevitability.
◈ The Linguistic Shift
Observe the transformation in the text:
- Instead of: "The Biohub applied this technology institutionally..."
- The text uses: "Institutional application of this technology has already manifested..."
By transforming the action (apply) into a noun phrase (institutional application), the author shifts the focus from the actor (the Biohub) to the phenomenon itself. This is the hallmark of C2 discourse: it removes the 'human' element to prioritize the 'systemic' result.
◈ Syntactic Sophistication: The 'Heavy' Subject
C2 mastery involves managing high cognitive loads via complex subject strings. Note the construction:
"The underlying architecture, ESMFold2, utilizes a protein language model trained on evolutionary data."
Here, the subject is not just a noun, but a defined entity qualified by an appositive. This allows the writer to embed a definition and a function into a single clause without breaking the flow—a technique known as Information Density.
◈ Lexical Precision for the High-End Academic
B2 students use effective; C2 practitioners use efficacy. B2 students use similarities; C2 practitioners use structural homologies.
The Critical Distinction:
- Homology Similarity.
- Similarity is a surface observation. Homology implies a shared evolutionary origin. This level of precision is what distinguishes a proficient speaker from a master of the language; the word choice does not just describe the object, it defines the scientific relationship of the object.
◈ The 'Open-Source' Collocation
Notice the phrase: "opted for an open-source distribution model."
Rather than saying "they decided to make it free," the text employs a collocational chain: opted for open-source distribution model. This creates a professional cadence that signals authority and institutional belonging.