Nature computational science最新文献_第5页

Structural alignments to design functional RNAs 结构比对设计功能性rna。

IF 20.3

Nature computational science Pub Date : 2026-07-30 DOI: 10.1038/s43588-026-01028-3

Gina El Nesr

引用次数: 0

Structure-alignment-driven cross-graph modeling for functional RNA design 功能性RNA设计的结构比对驱动交叉图建模。

IF 20.3

Nature computational science Pub Date : 2026-07-30 DOI: 10.1038/s43588-026-01029-2

Shengfan Wang, Jun Wang, Xiaojian Liu, Weimin Zhu, Jiaqi Huang, Yiming Xue, Junchi Yan, Hong-Bin Shen, Jie Song, Xiaoyong Pan

{"title":"Structure-alignment-driven cross-graph modeling for functional RNA design","authors":"Shengfan Wang, Jun Wang, Xiaojian Liu, Weimin Zhu, Jiaqi Huang, Yiming Xue, Junchi Yan, Hong-Bin Shen, Jie Song, Xiaoyong Pan","doi":"10.1038/s43588-026-01029-2","DOIUrl":"10.1038/s43588-026-01029-2","url":null,"abstract":"RNAs are critical for biological processes, and their functions are closely tied to their three-dimensional structures. The design of RNA sequences that fold into target three-dimensional structures is a complex challenge due to the dynamic and unstable nature of RNA structures. Here we present an RNA design approach, AlignIF, that leverages multiple structure alignment with cross-graph modeling and carefully hand-designed features to capture evolutionarily conserved structural patterns at the structural level, facilitating RNA sequence design. AlignIF outperforms existing state-of-the-art methods in test sets. Notably, it enables the design of entire RNA families rather than being restricted to recapitulating native sequences. Furthermore, AlignIF successfully designs functional RNA fluorescent aptamers and self-cleaving ribozymes that exhibit measurable activity without requiring any postdesign filtering. For Mango-I, two of ten designed aptamers showed enhanced fluorescence, and two others exhibited improved binding affinity compared with the wild type. For the more challenging iMango-III, all designs were fluorescent, including one with a sequence identity of ~30% relative to the wild type. The results highlight the potential of AlignIF for functional RNA engineering. This study presents AlignIF, which uses multiple structure alignments to capture evolutionary rules and conservation patterns across RNA families, enabling diverse, structure-guided designs of functional aptamers and ribozymes.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 8","pages":"826-841"},"PeriodicalIF":20.3,"publicationDate":"2026-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148633088","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

SpatialFormer: universal spatial representation learning from subcellular molecular to multicellular landscapes SpatialFormer：从亚细胞分子到多细胞景观的普遍空间表征学习。

IF 20.3

Nature computational science Pub Date : 2026-07-30 DOI: 10.1038/s43588-026-01016-7

Jun Wang, Yuanhua Huang, Ole Winther

{"title":"SpatialFormer: universal spatial representation learning from subcellular molecular to multicellular landscapes","authors":"Jun Wang, Yuanhua Huang, Ole Winther","doi":"10.1038/s43588-026-01016-7","DOIUrl":"10.1038/s43588-026-01016-7","url":null,"abstract":"Understanding gene spatial expression and the organization of multicellular systems is vital for disease diagnosis and studying biological processes. However, existing models often struggle to integrate gene expression data with cellular spatial information effectively. Here we introduce SpatialFormer, a hybrid framework combining convolutional networks and transformers to learn single-cell multimodal and multiscale information in the niche context, including expression data and subcellular gene spatial distribution. Pretrained on 700 million cell pairs from 17 million spatially resolved single cells across 71 Xenium slides, SpatialFormer merges gene spatial expression profiles with cell niche information via the pairwise training strategy. Our findings demonstrate that SpatialFormer distills biological signals across various tasks, including single-cell batch correction, cell-type annotation and co-localization detection. The perturbation analysis identified gene pairs essential for the immune cell–cell communication in pulmonary fibrosis, epithelial–myoepithelial co-localization and tumor transition signals in breast cancer. These advancements enhance our understanding of cellular dynamics and offer additional pathways for applications in biomedical research. SpatialFormer, a transformer-based model for spatial transcriptomics, learns cell–cell relationships to predict cell co-localization, annotate cell types and niches, and reveal cell communication gene pairs in pulmonary fibrosis.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 9","pages":"971-987"},"PeriodicalIF":20.3,"publicationDate":"2026-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148633165","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

Rethinking privacy-preserving GWAS benchmarks for governance 重新思考保护隐私的GWAS治理基准

IF 20.3

Nature computational science Pub Date : 2026-07-23 DOI: 10.1038/s43588-026-01023-8

Natnatee Dokmai

引用次数: 0

Cross-task, explainable and real-time decoding of human emotion states by integrating gray and white matter intracranial neural activity. 通过整合灰质和白质颅内神经活动对人类情绪状态进行跨任务、可解释和实时解码。

IF 20.3

Nature computational science Pub Date : 2026-07-22 DOI: 10.1038/s43588-026-01021-w

Yuxiao Yang, Wenjun Chen, Yuyang Chen, Ling Ding, Chi Zhang, Hongjie Jiang, Zhoule Zhu, Xinxia Guo, Shuang Wang, Gang Pan, Ning Wei, Shaohua Hu, Junming Zhu, Yueming Wang

{"title":"Cross-task, explainable and real-time decoding of human emotion states by integrating gray and white matter intracranial neural activity.","authors":"Yuxiao Yang, Wenjun Chen, Yuyang Chen, Ling Ding, Chi Zhang, Hongjie Jiang, Zhoule Zhu, Xinxia Guo, Shuang Wang, Gang Pan, Ning Wei, Shaohua Hu, Junming Zhu, Yueming Wang","doi":"10.1038/s43588-026-01021-w","DOIUrl":"10.1038/s43588-026-01021-w","url":null,"abstract":"<p><p>Decoding human emotion states from intracranial neural activity is key in developing affective brain-computer interfaces and new therapies for affective disorders. However, real-world application of decoding requires high performance that integrates neural activity from both gray and white matter, stable generalization across different contexts, sufficient neural encoding explainability and robust real-time implementation, all of which remain elusive. Here we simultaneously recorded intracranial electroencephalogram (iEEG) and abundant self-rated valence and arousal scores across two emotion-eliciting tasks in 18 individuals. We then developed personalized decoding models within a deep learning framework, achieving high-performance decoding of continuous valence and arousal states and improving on the performance of prior EEG and iEEG decoding. Critically, the models substantially improved performance by integrating gray and white matter signals and demonstrated cross-task generalization. The models further revealed shared and preferred mesolimbic-thalamo-cortical subnetworks encoding valence and arousal, showing neurophysiological explainability. Finally, the models realized robust real-time decoding in four new individuals. Our results have implications for advancing emotion decoding neurotechnology toward deployable affective brain-computer interfaces and closed-loop therapeutic systems for affective disorders.</p>","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":" ","pages":""},"PeriodicalIF":20.3,"publicationDate":"2026-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148563954","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

AlphaFold2 turns five AlphaFold2满五岁了。

IF 20.3

Nature computational science Pub Date : 2026-07-15 DOI: 10.1038/s43588-026-01031-8

引用次数: 0

Fast reconstruction of evolutionary trees from millions of genome sequences 从数百万个基因组序列中快速重建进化树。

IF 20.3

Nature computational science Pub Date : 2026-07-15 DOI: 10.1038/s43588-026-01014-9

引用次数: 0

UniFFBench: evaluating universal machine learning force fields against experimental measurements UniFFBench：根据实验测量评估通用机器学习力场。

IF 20.3

Nature computational science Pub Date : 2026-07-14 DOI: 10.1038/s43588-026-01019-4

Sajid Mannan, Vaibhav Bihani, Carmelo Gonzales, Kin Long Kelvin Lee, Nitya Nand Gosvami, Sayan Ranu, Santiago Miret, N. M. Anoop Krishnan

{"title":"UniFFBench: evaluating universal machine learning force fields against experimental measurements","authors":"Sajid Mannan, Vaibhav Bihani, Carmelo Gonzales, Kin Long Kelvin Lee, Nitya Nand Gosvami, Sayan Ranu, Santiago Miret, N. M. Anoop Krishnan","doi":"10.1038/s43588-026-01019-4","DOIUrl":"10.1038/s43588-026-01019-4","url":null,"abstract":"Universal machine learning force fields (UMLFFs) promise to revolutionize materials science by enabling rapid atomistic simulations across the periodic table. However, their evaluation has been limited to computational benchmarks that may not reflect real-world performance. Here we introduce UniFFBench, a comprehensive evaluation framework featuring the MinX dataset—a diverse collection of 1,500+ mineral systems spanning 85 elements, extreme thermodynamic conditions (0–5,000 K, 0–1,000 GPa) and structural complexity, including partial occupancy and disorder. This diversity, combined with experimental reference values for validation, enables assessment of UMLFF generalization across chemical space and conditions substantially beyond typical training scenarios. Our systematic evaluation of six state-of-the-art UMLFFs reveals a substantial ‘reality gap’: models achieving impressive performance on computational benchmarks often fail when confronted with experimental complexity. Even the best-performing models exhibit higher density prediction error than the threshold required for practical applications. We observe disconnects between simulation stability and mechanical property accuracy, with prediction errors correlating with training data representation rather than the modeling method. Universal machine learning force fields excel on computational benchmarks but consistently fail when tested against experimental measurements of real minerals, underscoring the need for evaluation frameworks grounded in real-world conditions and multiple material properties beyond energy and forces.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 8","pages":"894-906"},"PeriodicalIF":20.3,"publicationDate":"2026-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148450882","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

Generative model for information metamaterial design 信息超材料设计的生成模型。

IF 20.3

Nature computational science Pub Date : 2026-07-14 DOI: 10.1038/s43588-026-01025-6

Junming Hou, Long Chen, Xuan Zheng, Jia Wei Wu, Jian Wei You, Zi Xuan Cai, Jiahan Huang, Chenxu Wu, Jian Lin Su, Lianlin Li, Jia Nan Zhang, Tie Jun Cui

{"title":"Generative model for information metamaterial design","authors":"Junming Hou, Long Chen, Xuan Zheng, Jia Wei Wu, Jian Wei You, Zi Xuan Cai, Jiahan Huang, Chenxu Wu, Jian Lin Su, Lianlin Li, Jia Nan Zhang, Tie Jun Cui","doi":"10.1038/s43588-026-01025-6","DOIUrl":"10.1038/s43588-026-01025-6","url":null,"abstract":"Information metamaterials are digital coding electromagnetic structures that connect wave control with information processing, offering programmable routes to beam shaping, focusing and holographic imaging. Their inverse design remains challenging because both meta-atoms and spatial coding arrays must be selected from a vast combinatorial space, and existing optimization or learning methods are often tailored to specific tasks, bit resolutions or field patterns. Here we show a generative model for information metamaterial design that learns a shared design prior and transfers it across diverse electromagnetic functions. The model combines a pretrained diffusion backbone with lightweight functionality-oriented adapters, enabling the generation of multibit meta-atoms with prescribed responses and nonuniform arrays for beam steering, near-field focusing and holography. Numerical simulations and experiments validate high-performance meta-atoms and functional 1-bit and 3-bit meta-arrays. For holographic design, the model reaches Gerchberg–Saxton-level fidelity while reducing runtime by over three orders of magnitude, establishing a scalable route to information-metamaterial discovery. A generative model designs information metamaterials from meta-atoms to programmable arrays, enabling beam steering, focusing and holography with experimental validation and more than 1,000-fold faster holographic design.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 8","pages":"870-881"},"PeriodicalIF":20.3,"publicationDate":"2026-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148450829","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

Understanding language model scaling for protein fitness prediction 理解语言模型尺度对蛋白质适应度预测的影响。

IF 20.3

Nature computational science Pub Date : 2026-07-13 DOI: 10.1038/s43588-026-01010-z

Chao Hou, Di Liu, Aziz Zafar, Yufeng Shen

{"title":"Understanding language model scaling for protein fitness prediction","authors":"Chao Hou, Di Liu, Aziz Zafar, Yufeng Shen","doi":"10.1038/s43588-026-01010-z","DOIUrl":"10.1038/s43588-026-01010-z","url":null,"abstract":"Protein language models, as well as models that incorporate structural information or homologous sequences, estimate the sequence likelihood p(sequence), which reflects the protein fitness landscape and is commonly used in mutation effect prediction and protein design. It is widely believed in deep learning field that larger models perform better across tasks. However, for fitness prediction, language model performance declines beyond a certain size, raising concerns about their scalability. Here we showed that model size, training dataset and stochastic elements can bias the predicted p(sequence) away from real fitness. Model performance on fitness prediction depends on how well p(sequence) matches evolutionary patterns in homologs, which is best achieved at a moderate p(sequence) level for most proteins. At extreme predicted wild-type sequence likelihoods, models predict uniformly low or high likelihoods for nearly all mutations, failing to reflect the real fitness landscape. Notably, larger models tend to predict proteins with higher p(sequence), which may exceed the moderate range and thus reduce performance. Our findings clarify the scaling behavior of protein models on fitness prediction and provide practical guidelines for their application and future development. Larger language models do not always perform better at fitness prediction. Hou et al. found that performance depends on whether model outputs match evolutionary patterns in homologs, which is best achieved at moderate predicted sequence likelihoods.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 7","pages":"778-788"},"PeriodicalIF":20.3,"publicationDate":"2026-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148439124","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