编译|李言

Science,  30 April 2026, Volume 392 Issue 6797

《科学》2026年4月30日，第392卷，6797期

化学Chemistry

Precision indole skeletal editing for single-carbon replacement

精准编辑吲哚骨架实现单碳置换

▲ 作者：Ling Zhang, Yatao Lang, Zhen Luo, Xinlong Han et al.

▲链接：

https://www.science.org/doi/10.1126/science.aec3587

▲摘要：

吲哚广泛存在于药物和天然产物中。研究者展示了色胺衍生物的一种分子内骨架编辑反应。该反应通过侧链酰胺的光反应实现，同时完成了吲哚C2位区域选择性的单碳置换与取代。该策略不仅能够在吲哚C2位实现氘代、烷基化、芳基化和酰基化，还能够将13C标记的碳原子引入骨架中。

通过对复杂单萜吲哚生物碱quebrachamine的简洁四步全合成，研究者验证了该方法的实用价值。实验研究与密度泛函理论计算揭示了一条涉及级联[2+2]环加成、retro-[2+2]开环、脱羰基及环化的反应路径，从原子层面阐明了反应历程。性能稳定，且得益于电缆聚合物护套的防潮保护，仍能保留80%的导电性。

▲ Abstract：

Indole is widely present in pharmaceuticals and natural products. In this work, we report an intramolecular skeletal editing reaction of tryptamine derivatives enabled by photoreaction of a pendant amide that simultaneously achieves regioselective single-carbon replacement and substitution at the C2 position of indoles. This strategy facilitates deuteration, alkylation, arylation, and acylation at the indole C2 position while also enabling the incorporation of 13C-labeled carbon into the scaffold. We highlighted the practical applicability of this method through a concise four-step total synthesis of quebrachamine, a complex monoterpene indole alkaloid. Experimental studies and density functional theory (DFT) calculations revealed a reaction pathway involving cascade [2+2] cycloaddition, retro-[2+2] ring opening, decarbonylation, and cyclization, elucidating the sequence at an atomic level.

人工智能Artificial Intelligence

Toward life with a 19–amino acid alphabet through generative artificial intelligence design

通过生成式人工智能设计迈向含19种氨基酸字母表的生命体系

▲ 作者：Liyuan Liu, Charlotte Rochereau et al.

▲链接：

https://www.science.org/doi/10.1126/science.aeb5171

▲摘要：

由于所有已知生物体均由至少20种经典氨基酸构成，使用更简化的氨基酸“字母表”的可行性尚不明确。研究者结合计算设计与合成生物学，探索从含19种氨基酸字母表构建细胞的可能性。

初步分析表明，异亮氨酸可能是可被剔除的，研究者通过直接替换大肠杆菌必需蛋白中的所有异亮氨酸残基证实了这一点。关键的是，在大多数情况下，需要利用蛋白质语言模型和基于结构的模型来重新设计无异亮氨酸的功能性蛋白。

研究者系统性地替换了核糖体中全部382个异亮氨酸残基，并在原生基因组位点上组合了21个重新设计的亚基，成功获得了可存活且进化稳定的细胞。这项工作为自早期进化以来首个含19种氨基酸生物体的构建提供了路线图。

▲ Abstract：

Because all known living organisms are made from at least 20 canonical amino acids, the feasibility of life using a more simplified alphabet remains unclear. In this work, we leveraged computational design and synthetic biology to explore building a cell from a 19–amino acid alphabet. Initial analyses suggested that isoleucine (Ile) may be dispensable, which we confirmed by directly replacing Ile residues in essential proteins in Escherichia coli. Critically, protein language models and structure-based models were necessary to redesign functional Ile-less proteins in most cases. We systematically replaced all 382 Ile residues from the ribosome and combined 21 redesigned subunits at a native genomic locus to produce a viable, evolutionarily stable cell. This work provides a roadmap to create the first 19–amino acid organism since early evolution.

Performance of a large language model on the reasoning tasks of a physician

大语言模型在医疗诊断任务中的表现

▲ 作者：Peter G. Brodeur, Thomas A. Buckley et al.

▲链接：

https://www.science.org/doi/10.1126/science.adz4433

▲摘要：

至少65年以前，复杂的临床诊断推理案例就已被引入作为评估专家医学计算系统的金标准，这一标准沿用至今。研究者通过五项实验，以数百名医师为基线，报告了对大型语言模型在具有挑战性的临床案例上的医师评估结果。

随后，研究者报告了一项真实世界研究，在某大型三级学术医学中心的急诊科，随机选取患者，比较人类专家与人工智能第二意见的诊断表现。在所有实验中，大语言模型均超越医师基线水平，并展现出相较于前几代AI临床决策支持系统的持续改进。

研究表明，大语言模型已超越大多数临床推理基准，亟需推进前瞻性临床试验。

▲ Abstract：

More than 65 years ago, complex clinical diagnostic reasoning cases were introduced as the gold standard for the evaluation of expert medical computing systems, a standard that has held ever since. In this study, we report the results of a physician evaluation of a large language model (LLM) on challenging clinical cases across five experiments with a baseline of hundreds of physicians. We then report a real-world study comparing human expert and artificial intelligence (AI) second opinions in randomly selected patients in the emergency room of a major tertiary academic medical center. In all experiments, the LLM outperformed physician baselines and displayed continued improvement from prior generations of AI clinical decision support. Our study suggests that LLMs have eclipsed most benchmarks of clinical reasoning, motivating the urgent need for prospective trials.

天文学Astronomy

Deeper detection limits in astronomical imaging using self-supervised spatiotemporal denoising

自监督时空降噪提升天文成像探测极限

▲ 作者：Yuduo Guo, Hao Zhang et al.

▲链接：

https://www.science.org/doi/10.1126/science.ady9404

▲摘要：

天文成像观测的探测极限受多种噪声源制约。其中部分噪声在相邻像素与曝光序列间存在相关性，因此理论上可以通过学习进行校正。研究者提出了基于自监督Transformer的天文去噪算法，该算法整合了多幅曝光图像中的时空信息。

对模拟数据的基准测试表明，ASTERIS在保持点扩展函数和测光精度的前提下，在90%完备性和纯净度条件下将探测极限提升了1.0星等。利用詹姆斯·韦伯空间望远镜和斯巴鲁望远镜数据的观测验证，识别出此前无法探测的特征，包括低表面亮度星系结构和引力透镜弧。

将ASTERIS应用于韦伯望远镜的深场图像后，红移?9的星系候选体数量较传统方法增加了三倍，其静止帧紫外光度暗至1.0星等。

▲ Abstract：

The detection limit of astronomical imaging observations is limited by several noise sources. Some of that noise is correlated between neighboring pixels and exposures, so in principle it could be learned and corrected. We present the Astronomical Self-supervised Transformer-based Denoising (ASTERIS) algorithm, which integrates spatiotemporal information across multiple exposures. Benchmarking on mock data indicated that ASTERIS improves detection limits by 1.0 magnitude at 90% completeness and purity while preserving the point spread function and photometric accuracy. Observational validation using data from the James Webb Space Telescope (JWST) and the Subaru Telescope identified previously undetectable features, including low-surface-brightness galaxy structures and gravitationally lensed arcs. Applied to deep JWST images, ASTERIS identified three times more redshift ?9 galaxy candidates than previous methods, with rest-frame ultraviolet luminosity 1.0 magnitude fainter.

地球科学Earth Science

Apparent Hack’s law in river deltas

河流三角洲中显见的哈克定律

▲ 作者：Tian Y. Dong, Lawrence Vulis et al.

▲链接：

https://www.science.org/doi/10.1126/science.ady6805

▲摘要：

河流三角洲是人口密集、生态至关重要的地貌系统，正面临海平面上升的威胁。分流河道网络通过输送沉积物构建三角洲陆地，但网络组织与陆地构建之间的关系仍不清楚。

受哈克定律（该定律表明流域汇水面积与支流网络中的河道长度呈标度关系）启发，研究者分析了一套全球分流网络数据集，发现分流河道长度与其对应的陆地构建面积（相当于汇水区中的汇水面积）之间呈现近乎相同的标度关系。

尽管存在这种全球性标度规律，研究者进一步识别出两种不同的局地造陆模式：均匀三角洲网络始终遵循哈克定律，而复合三角洲网络则呈现标度断裂，即从三角洲顶点附近的充填式生长过渡到海岸附近的准线性生长。这些出乎意料的增长模式表明，河流三角洲的生长与组织过程中，全球规律性与局地多样性是共存的。

▲ Abstract：

River deltas are densely populated, ecologically vital landscapes threatened by rising sea levels. Distributary channel networks disperse sediment to build deltaic land, yet the relationship between the network organization and land building remains elusive. Inspired by Hack’s law, which shows that watershed drainage area scales with channel length in tributary networks, we analyzed a global dataset of distributary networks and found a nearly identical scaling relationship between distributary channel length and nourishment area, the land-building counterpart to drainage area. Despite this apparent global scaling, we further identified two distinct local land-building patterns: uniform delta networks consistently follow Hack’s law, whereas composite delta networks exhibit a scale break, transitioning from space-filling growth around the delta apex to quasi-linear growth near the coast. The unexpected growth patterns suggest that global simplicity and local variability coexist in how river deltas grow and organize.

网络科学Network Science

Peer influence decay and behavioral diffusion in adolescent networks: A simulation approach

青少年网络中同伴影响力衰减与行为扩散：一项模拟研究

▲ 作者：Cheng Wang, Carter T. Butts et al.

▲链接：

https://www.science.org/doi/10.1126/science.aea9297

▲摘要：

同伴影响力在社交网络中传播多远后会消散？研究者利用美国国家青少年至成人健康纵向研究中两所学校（共3154名学生）的纵向数据，探讨了青少年友谊网络中吸烟行为的扩散机制。

研究者采用随机面向行为者模型，模拟了针对重度吸烟者的干预措施，涉及不同策略（随机选择、入度中心性、特征向量中心性）与覆盖比例（10%至100%）。通过一个新的指数衰减模型量化了影响力衰减，首次揭示了目标个体三阶以内的间接同伴影响力（即溢出效应）。对10%至30%的核心个体实施干预可使吸烟行为降幅最大化，但由于网络饱和效应，当覆盖比例超过40%—50%后增益趋于平缓。

研究分析表明，相较于更大规模但更稀疏的网络，较密集的网络表现出更广泛的扩散范围与更缓慢的衰减速率。该衰减指标可为不同结构网络中的干预方案优化提供依据。

▲ Abstract：

How far does peer influence spread through social networks before dissipating? This study investigates the diffusion of smoking behavior in adolescent friendship networks using longitudinal data from two schools (n = 3154 students) in the National Longitudinal Study of Adolescent to Adult Health. Using Stochastic Actor–Oriented Models, we simulate interventions targeting heavy smokers using various strategies (random, in-degree, eigenvector centrality) and coverage (10 to 100%). A new exponential decay model quantifies influence attenuation, revealing indirect peer influences, or spillover effects, up to three steps from targets. Targeting 10 to 30% of central individuals maximizes smoking reductions, but gains plateau beyond 40 to 50% owing to network saturation. In our analyses, the denser network exhibits broader diffusion and slower decay than the larger, sparser network. This decay metric optimizes intervention design across diverse network structures.