Nature computational science最新文献_第2页

Analog in-memory computing attention mechanism for fast and energy-efficient large language models. 模拟内存计算关注机制的快速和节能的大型语言模型。

IF 18.3

Nature computational science Pub Date : 2025-09-08 DOI: 10.1038/s43588-025-00854-1

Nathan Leroux, Paul-Philipp Manea, Chirag Sudarshan, Jan Finkbeiner, Sebastian Siegel, John Paul Strachan, Emre Neftci

{"title":"Analog in-memory computing attention mechanism for fast and energy-efficient large language models.","authors":"Nathan Leroux, Paul-Philipp Manea, Chirag Sudarshan, Jan Finkbeiner, Sebastian Siegel, John Paul Strachan, Emre Neftci","doi":"10.1038/s43588-025-00854-1","DOIUrl":"https://doi.org/10.1038/s43588-025-00854-1","url":null,"abstract":"Transformer networks, driven by self-attention, are central to large language models. In generative transformers, self-attention uses cache memory to store token projections, avoiding recomputation at each time step. However, graphics processing unit (GPU)-stored projections must be loaded into static random-access memory for each new generation step, causing latency and energy bottlenecks. Here we present a custom self-attention in-memory computing architecture based on emerging charge-based memories called gain cells, which can be efficiently written to store new tokens during sequence generation and enable parallel analog dot-product computation required for self-attention. However, the analog gain-cell circuits introduce non-idealities and constraints preventing the direct mapping of pre-trained models. To circumvent this problem, we design an initialization algorithm achieving text-processing performance comparable to GPT-2 without training from scratch. Our architecture reduces attention latency and energy consumption by up to two and four orders of magnitude, respectively, compared with GPUs, marking a substantial step toward ultrafast, low-power generative transformers.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":18.3,"publicationDate":"2025-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145024889","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A digital twin that interprets and refines chemical mechanisms. 一个解释和完善化学机制的数字双胞胎。

IF 18.3

Nature computational science Pub Date : 2025-09-01 DOI: 10.1038/s43588-025-00859-w

引用次数: 0

MetaKSSD: boosting the scalability of the reference taxonomic marker database and the performance of metagenomic profiling using sketch operations. MetaKSSD：提高参考分类标记数据库的可扩展性和使用草图操作的宏基因组分析的性能。

IF 18.3

Nature computational science Pub Date : 2025-08-29 DOI: 10.1038/s43588-025-00855-0

Huiguang Yi, Xiaoxin Lu, Qing Chang

{"title":"MetaKSSD: boosting the scalability of the reference taxonomic marker database and the performance of metagenomic profiling using sketch operations.","authors":"Huiguang Yi, Xiaoxin Lu, Qing Chang","doi":"10.1038/s43588-025-00855-0","DOIUrl":"https://doi.org/10.1038/s43588-025-00855-0","url":null,"abstract":"The performance of metagenomic profiling is constrained by the diversity of taxa present in the reference taxonomic marker database (MarkerDB) used. However, continually updating MarkerDB to include newly determined taxa using existing approaches faces increasing difficulties and will soon become impractical. Here we introduce MetaKSSD, which redefines MarkerDB construction and metagenomic profiling using sketch operations, enhancing MarkerDB scalability and profiling performance. MetaKSSD encompasses 85,202 species in its MarkerDB using just 0.17 GB of storage and profiles 10 GB of data within seconds. Leveraging its comprehensive MarkerDB, MetaKSSD substantially improves profiling results. In a microbiome-phenotype association study, MetaKSSD identified more effective associations than MetaPhlAn4. We profiled 382,016 metagenomic runs using MetaKSSD, conducted extensive sample clustering analyses and suggested potential yet-to-be-discovered niches. MetaKSSD offers functionality for instantaneous searching of similar profiles. It enables the swift transmission of metagenome sketches over the network and real-time online metagenomic analysis, facilitating use by non-expert users.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":18.3,"publicationDate":"2025-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144981670","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Digital Twin for Chemical Science: a case study on water interactions on the Ag(111) surface. 化学科学的数字孪生：Ag（111）表面水相互作用的案例研究。

IF 18.3

Nature computational science Pub Date : 2025-08-27 DOI: 10.1038/s43588-025-00857-y

Jin Qian, Asmita Jana, Siddarth Menon, Andrew E Bogdan, Rebecca Hamlyn, Johannes Mahl, Ethan J Crumlin

{"title":"Digital Twin for Chemical Science: a case study on water interactions on the Ag(111) surface.","authors":"Jin Qian, Asmita Jana, Siddarth Menon, Andrew E Bogdan, Rebecca Hamlyn, Johannes Mahl, Ethan J Crumlin","doi":"10.1038/s43588-025-00857-y","DOIUrl":"10.1038/s43588-025-00857-y","url":null,"abstract":"Directly visualizing chemical trajectories offers insights into catalysis, gas-phase reactions and photoinduced dynamics. Tracking the transformation of chemical species is best achieved by coupling theory and experiment. Here we developed Digital Twin for Chemical Science (DTCS) v.01, which integrates theory, experiment and their bidirectional feedback loops into a unified platform for chemical characterization. DTCS addresses a core question: given a set of experimental conditions, what is the expected outcome and why? It consists of a forward solver that takes a chemical reaction network and predicts spectra under experimental conditions, and an inverse solver that infers kinetics from measured spectra. We applied DTCS to ambient-pressure X-ray photoelectron spectroscopy measurements of the Ag-H2O interface as an example. This approach enables real-time knowledge extraction and guides experiments until a stopping condition is met based on accuracy and degeneracy. As a step toward autonomous chemical characterization, DTCS provides mechanistic knowledge in a verified, standardized manner.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":18.3,"publicationDate":"2025-08-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144981653","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Pharmacophore-oriented 3D molecular generation toward efficient feature-customized drug discovery. 面向药物团的三维分子生成，实现高效的特征定制药物发现。

IF 18.3

Nature computational science Pub Date : 2025-08-26 DOI: 10.1038/s43588-025-00850-5

Jian Peng, Jun-Lin Yu, Zeng-Bao Yang, Yi-Ting Chen, Si-Qi Wei, Fan-Bo Meng, Yao-Geng Wang, Xiao-Tian Huang, Guo-Bo Li

{"title":"Pharmacophore-oriented 3D molecular generation toward efficient feature-customized drug discovery.","authors":"Jian Peng, Jun-Lin Yu, Zeng-Bao Yang, Yi-Ting Chen, Si-Qi Wei, Fan-Bo Meng, Yao-Geng Wang, Xiao-Tian Huang, Guo-Bo Li","doi":"10.1038/s43588-025-00850-5","DOIUrl":"https://doi.org/10.1038/s43588-025-00850-5","url":null,"abstract":"Molecular generation is a cutting-edge technology with the potential to revolutionize intelligent drug discovery. However, currently reported ligand-based or structure-based molecular generation methods remain unpractical for real-world drug discovery. Here we propose an explicit pharmacophore-oriented 3D molecular generation method, termed PhoreGen. PhoreGen employs asynchronous perturbations and updates on both atomic and bond information, coupled with a message-passing mechanism that incorporates prior knowledge of ligand-pharmacophore mapping during the diffusion-denoising process. Evaluations revealed that PhoreGen efficiently generates 3D molecules well aligned with pharmacophores, maintaining good chemical reasonability, diversity, drug-likeness and binding affinity and, importantly, produces feature-customized molecules at high frequency. By using PhoreGen, we successfully identified new bicyclic boronate inhibitors of evolved metallo-β-lactamase and serine-β-lactamases, which potentiate meropenem against clinically isolated superbugs. Moreover, we identified inhibitors of metallo-nicotinamidases, emerging targets for insecticides. This work explores an explicitly constrained mode for molecular generation and demonstrates its potential in feature-customized drug discovery.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":18.3,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144981722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep active optimization for complex systems. 复杂系统的深度主动优化。

IF 18.3

Nature computational science Pub Date : 2025-08-25 DOI: 10.1038/s43588-025-00858-x

Ye Wei, Bo Peng, Ruiwen Xie, Yangtao Chen, Yu Qin, Peng Wen, Stefan Bauer, Po-Yen Tung, Dierk Raabe

{"title":"Deep active optimization for complex systems.","authors":"Ye Wei, Bo Peng, Ruiwen Xie, Yangtao Chen, Yu Qin, Peng Wen, Stefan Bauer, Po-Yen Tung, Dierk Raabe","doi":"10.1038/s43588-025-00858-x","DOIUrl":"https://doi.org/10.1038/s43588-025-00858-x","url":null,"abstract":"Inferring optimal solutions from limited data is considered the ultimate goal in scientific discovery. Artificial intelligence offers a promising avenue to greatly accelerate this process. Existing methods often depend on large datasets, strong assumptions about objective functions, and classic machine learning techniques, restricting their effectiveness to low-dimensional or data-rich problems. Here we introduce an optimization pipeline that can effectively tackle complex, high-dimensional problems with limited data. This approach utilizes a deep neural surrogate to iteratively find optimal solutions and introduces additional mechanisms to avoid local optima, thereby minimizing the required samples. Our method finds superior solutions in problems with up to 2,000 dimensions, whereas existing approaches are confined to 100 dimensions and need considerably more data. It excels across varied real-world systems, outperforming current algorithms and enabling efficient knowledge discovery. Although focused on scientific problems, its benefits extend to numerous quantitative fields, paving the way for advanced self-driving laboratories.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":18.3,"publicationDate":"2025-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144981709","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Exploiting pleiotropy to enhance variant discovery with functional false discovery rates. 利用多效性提高变异发现与功能性错误发现率。

IF 18.3

Nature computational science Pub Date : 2025-08-22 DOI: 10.1038/s43588-025-00852-3

Andrew J Bass, Chris Wallace

引用次数: 0

Author Correction: Probing the limitations of multimodal language models for chemistry and materials research. 作者更正：探讨化学和材料研究中多模态语言模型的局限性。

IF 18.3

Nature computational science Pub Date : 2025-08-21 DOI: 10.1038/s43588-025-00869-8

Nawaf Alampara, Mara Schilling-Wilhelmi, Martiño Ríos-García, Indrajeet Mandal, Pranav Khetarpal, Hargun Singh Grover, N M Anoop Krishnan, Kevin Maik Jablonka

引用次数: 0

SciToolAgent: a knowledge-graph-driven scientific agent for multitool integration. SciToolAgent：知识图驱动的科学代理，用于多工具集成。

IF 18.3

Nature computational science Pub Date : 2025-08-20 DOI: 10.1038/s43588-025-00849-y

Keyan Ding, Jing Yu, Junjie Huang, Yuchen Yang, Qiang Zhang, Huajun Chen

{"title":"SciToolAgent: a knowledge-graph-driven scientific agent for multitool integration.","authors":"Keyan Ding, Jing Yu, Junjie Huang, Yuchen Yang, Qiang Zhang, Huajun Chen","doi":"10.1038/s43588-025-00849-y","DOIUrl":"https://doi.org/10.1038/s43588-025-00849-y","url":null,"abstract":"Scientific research increasingly relies on specialized computational tools, yet effectively utilizing these tools requires substantial domain expertise. While large language models show promise in tool automation, they struggle to seamlessly integrate and orchestrate multiple tools for complex scientific workflows. Here we present SciToolAgent, a large language model-powered agent that automates hundreds of scientific tools across biology, chemistry and materials science. At its core, SciToolAgent leverages a scientific tool knowledge graph that enables intelligent tool selection and execution through graph-based retrieval-augmented generation. The agent also incorporates a comprehensive safety-checking module to ensure responsible and ethical tool usage. Extensive evaluations on a curated benchmark demonstrate that SciToolAgent outperforms existing approaches. Case studies in protein engineering, chemical reactivity prediction, chemical synthesis and metal-organic framework screening further demonstrate SciToolAgent's capability to automate complex scientific workflows, making advanced research tools accessible to both experts and nonexperts.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":18.3,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144981676","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Engineering morphogenesis of cell clusters with differentiable programming. 可微规划细胞簇的工程形态发生。

IF 18.3

Nature computational science Pub Date : 2025-08-13 DOI: 10.1038/s43588-025-00851-4

Ramya Deshpande, Francesco Mottes, Ariana-Dalia Vlad, Michael P Brenner, Alma Dal Co

引用次数: 0