编译|冯维维

Science, 22 OCT 2026,VOLUME 394, ISSUE 6822

《科学》2026年10月22日，第394卷，6822期

生命科学Life Science

Wireless electrocatalytic closed-loop system for drug-free management of diabesity

无线电催化闭环系统用于无药物治疗糖胖病

▲ 作者：DEBASIS MAITY, PREETAM GUHA RAY, PETER BUCHMANN, GHISLAINE CHARPIN-EL HAMRI, AND MARTIN FUSSENEGGER

▲链接：

https://www.science.org/doi/10.1126/science.ady8560

▲摘要：

糖尿病和肥胖症在世界各地日益普遍，且二者常共存并相互加重。目前已存在多种药物和生物学干预手段，但各有优劣，许多患者仍需要耐受性更好、更有效和/或更可负担的治疗选择。

研究者开发的一种植入式装置有助于填补这一空白，它提供了一种无需药物或胰岛素即可可控调节血糖浓度的方法。

该无线供电装置可与临床可用的连续血糖监测仪连接，催化葡萄糖转化为葡萄糖酸盐，后者随后经肾脏排泄。在小鼠实验中，该装置使血糖恢复正常，同时去除了部分多余的热量负荷。

▲ Abstract：

Diabetes and obesity are becoming increasingly common around the world, and they frequently coexist and reinforce each other. A variety of pharmacological and biological interventions are already available, but each has its advantages and disadvantages and many patients are still in need of better tolerated, more effective, and/or more affordable treatment options. An implantable device developed by Maity et al. helps to fill this gap by providing a controllable way to adjust blood glucose concentration without medications or insulin (see the Perspective by Bentley and Payne). The wirelessly powered device can be linked to a clinically available continuous glucose monitor to catalyze the conversion of glucose to gluconate, which is then renally excreted. This normalized blood glucose and simultaneously removed some of the excess caloric load when tested in mice.

Noncoding transcription controls the developmental dynamics of long-range gene regulation

非编码转录控制长程基因调控的发育动态

▲ 作者：PHILIPPE J. BATUT AND MICHAEL S. LEVINE

▲链接：

https://www.science.org/doi/10.1126/science.aek5696

▲摘要：

调控基因表达的基因组区域本身也常被转录成多种非编码RNA（ncRNA）。然而，这种非编码转录的调控作用在很大程度上仍不清楚。

通过活体成像，研究者揭示了来自不同调控元件的ncRNA的序贯转录是果蝇胚胎中基因激活的基础。

单等位基因共可视化发现，最佳基因激活仅通过中等水平的增强子活性即可实现。破坏增强子相关ncRNA会导致基因过早激活，这提供了ncRNA控制发育中基因表达时序的证据。进一步研究表明，增强子转录可以调控复杂调控景观内的长程相互作用。

研究者提出，ncRNA在顺式作用下局部调节调控元件活性，从而塑造基因组组织并编排发育中基因表达的时间控制。

▲ Abstract：

The genomic regions regulating gene expression are often themselves transcribed into a variety of noncoding RNAs (ncRNAs). However, the regulatory roles of this noncoding transcription remain largely unknown. By using live imaging, we reveal that the sequential transcription of ncRNAs emanating from distinct regulatory elements underlies gene activation in Drosophila embryos. Single-allele covisualization uncovers that optimal gene activation is achieved by only moderate levels of enhancer activity. Disrupting enhancer-associated ncRNAs causes precocious gene activation, providing evidence that ncRNAs control the timing of gene expression in development. We further show that enhancer transcription can regulate long-range interactions within complex regulatory landscapes. We propose that ncRNAs locally modulate regulatory element activity in cis to shape genome organization and orchestrate the temporal control of gene expression in development.

人工智能Artificial Intellegence

The Virtual Biotech: A multi-agent AI framework for therapeutic discovery and development

虚拟生物技术公司：用于治疗发现与开发的多智能体AI框架

▲ 作者：HARRISON G. ZHANG, PETER ECKMANN, JIACHENG MIAO, ANDREW B. MAHON, AND JAMES ZOU

▲链接：

https://www.science.org/doi/10.1126/science.aeg6779

▲摘要：

传统药物研发是一项劳动密集且成本高昂的工作，不同研发环节需要独立的人员团队来完成。研究者开发了“虚拟生物技术公司（Virtual Biotech）”技术，旨在简化药物早期研发流程。

该技术由多个专业化AI智能体组成一套集成工作流，全部由一名充当首席科学官角色的AI智能体统一调度。和所有AI驱动的生物学研究一样，智能体给出的建议还需要通过合成候选药物、开展实验测试，才能落地到真实的现实研究中。

该方法仅适用于已有充足背景数据可供AI解读的疾病，但它有助于挖掘那些原本可能被人们忽略的治疗研发机会。

▲ Abstract：

Conventional drug discovery is a labor-intensive and expensive process, requiring separate teams of people for different steps. The Virtual Biotech approach was designed by Zhang et al. to streamline early drug discovery, with an integrated workflow made up of teams of specialized AI agents, all directed by an agent acting as a chief scientific officer. As with any AI-driven biological research, the agents’ recommendations will need to be translated into the real world by synthesizing the candidate drugs and testing them. This approach can only work for diseases for which there are enough background data for the AI to interpret, but it may help to identify therapeutic opportunities that would otherwise be missed.

Sequential diagnosis with language models

利用语言模型进行序贯诊断

▲ 作者：HARSHA NORI, MAYANK DASWANI, CHRISTOPHER KELLY, SCOTT LUNDBERG, MARCO TULIO RIBEIRO, MARC WILSON, XIAOXUAN LIU, VIKNESH SOUNDERAJAH, JONATHAN M. CARLSON, AND ERIC HORVITZ

▲链接：

https://www.science.org/doi/10.1126/science.aeh7930

▲摘要：

研究者探索了前沿语言模型在序贯诊断中的能力。序贯诊断是一种捕捉临床推理迭代本质的诊断范式。

给定简短的临床表现，临床医生有选择地获取信息以达成诊断，从而可以评估诊断准确性和检查成本。他们开发了一个交互式基准和一个AI系统，该系统模拟一组医生，不断修正鉴别诊断并选择具有成本效益的检查。

使用该基准，该系统在《新英格兰医学杂志》临床病理讨论会病例上达到83.2%的准确率，在MIMIC-IV-ED急诊科入院病例上达到80.3%，优于全科医生（分别为36.1%和46.8%）。

研究者表示，尽管这些发现是在受控的基准环境中获得的，但其凸显了语言模型在支持更准确、更低成本诊断方面的前景。

▲ Abstract：

We explored the capabilities of frontier language models in sequential diagnosis, a diagnostic paradigm that captures the iterative nature of clinical reasoning. Given a brief clinical presentation, clinicians selectively acquired information to reach a diagnosis, enabling assessment of diagnostic accuracy and workup cost. We developed an interactive benchmark and an artificial intelligence (AI) system that simulates a panel of physicians, continually revising the differential diagnosis and selecting cost-effective tests. Using this benchmark, the system achieved 83.2% accuracy on New England Journal of Medicine clinicopathological conference cases and 80.3% on MIMIC-IV-ED emergency department admissions, outperforming generalist physicians (36.1 and 46.8%, respectively). Although these findings were obtained in a controlled benchmark setting, they highlight the promise of language models for supporting more accurate, lower-cost diagnoses.

物理学Physics

Electronic bound states in the continuum in a two-dimensional metal

二维金属中的电子连续谱束缚态

▲ 作者：MORGAN THINEL, XIONG HUANG, DANIEL G. CHICA, SIMON TURKEL, SEBASTIAN E. ROSSI, CHRISTIE S. KOAY, HONG LI, WILLA MIHALYI-KOCH, BIPASA SAMANTA, AND RAQUEL QUEIROZ

▲链接：

https://www.science.org/doi/10.1126/science.adt9825

▲摘要：

尽管位于延展、离域态的连续谱之内，连续谱束缚态（BIC）却仍保持局域的量子态。光学BIC已在光子系统中实现，而在金属背景中实现电子束缚态仍是一项持续存在的挑战。

研究者利用扫描隧道显微镜识别出两种缺陷诱导的电子态。它们在Pd₅AlI₂（一种二维范德华金属）的金属连续谱内保持局域。

计算表明，Pd₅AlI₂装饰性棋盘格晶格上的电子跳跃干涉产生了一个（准）平带，该平带由与色散连续谱正交的空间局域本征态构成。这一阻挫晶格内的原子缺陷将相关的平带局域态移入金属连续谱，形成稳健的、对称性保护的BIC。

▲ Abstract：

Bound states in the continuum (BICs) are quantum states that remain localized despite lying within a continuum of extended, delocalized states. Whereas optical BICs have been realized in photonic systems, achieving electronic bound states embedded in a metallic background remains an ongoing challenge. In this study, we used scanning tunneling microscopy to identify two defect-induced electronic states that remain localized within the metallic continuum of Pd₅AlI₂, a two-dimensional van der Waals metal. Our calculations suggest that electron-hopping interference on the decorated checkerboard lattice of Pd₅AlI₂ gives rise to a (quasi) flat band built from spatially localized eigenstates that are orthogonal to the dispersive continuum. Atomic defects within this frustrated lattice shift the associated flat-band localized states into the metallic continuum, forming robust, symmetry-protected BICs.

Visualizing symmetry-broken Chern insulators and their quantum melting

可视化对称破缺的陈绝缘体及其量子熔化

▲ 作者：MINHAO HE, YEN-CHEN TSUI, RAN PENG, KENJI WATANABE, TAKASHI TANIGUCHI, OSKAR VAFEK, AND ALI YAZDANI

▲链接：

https://www.science.org/doi/10.1126/science.aef0940

▲摘要：

在磁场存在下，莫尔量子材料的电子态会形成霍夫施塔特谱，为研究能带拓扑与强电子-电子相互作用之间的相互作用提供了独特的环境。

利用扫描隧道显微镜，本文研究了与六方氮化硼对齐的双层石墨烯中的霍夫施塔特态，并直接可视化了相互作用驱动的对称破缺陈绝缘体的形成。

这一的测量揭示了在霍夫施塔特带的分数量填充下，形成使莫尔单胞加倍、三倍或四倍的物相，以及具有复杂单胞内波函数的态。

研究者在不同的陈态中可视化了两种不同的量子现象，包括由拓扑缺陷的出现和增殖驱动的量子熔化，以及与相竞争和相分离共存的量子相变。

▲ Abstract：

In the presence of a magnetic field, electronic states of moiré quantum materials develop a Hofstadter spectrum that provides a distinctive setting for studying the interplay between band topology and strong electron-electron interaction. Using scanning tunneling microscopy, we studied Hofstadter’s states in bilayer graphene aligned with hexagonal boron nitride and directly visualized the formation of interaction-driven symmetry breaking Chern insulators. Our measurements revealed the formation of phases that double, triple, or quadruple the moiré unit cell at fractional filling of the Hofstadter bands as well as states with complex intra–unit cell wave functions. We visualized two distinct quantum phenomena in different Chern states, including quantum melting driven by the appearance and proliferation of topological defects and a quantum transition co-occurring with phase competition and separation.