Nature computational science最新文献_第10页

The evolution of gene regulatory networks 基因调控网络的进化

IF 18.3

Nature computational science Pub Date : 2026-04-28 DOI: 10.1038/s43588-026-00981-3

Tatiana Belova, Daniel Osorio, Mariike L. Kuijjer

引用次数: 0

Sustainability as a challenge in complex systems dynamics 可持续性是复杂系统动力学中的一个挑战

IF 18.3

Nature computational science Pub Date : 2026-04-23 DOI: 10.1038/s43588-026-00971-5

Mehrnaz Anvari, Marc Timme

引用次数: 0

Enhancing success rates in therapeutic antibody design through generative models 通过生成模型提高治疗性抗体设计的成功率。

IF 20.3

Nature computational science Pub Date : 2026-04-17 DOI: 10.1038/s43588-026-00982-2

引用次数: 0

Feature-preserving manifold approximation and projection to analyze single-cell data 保持特征的流形逼近与投影分析单细胞数据。

IF 18.3

Nature computational science Pub Date : 2026-04-17 DOI: 10.1038/s43588-026-00970-6

Yang Yang, Jialei Gong, Hongjian Sun, Amos Choo, Jessica Cara Mar, Yunbo Wei, Yu Zhang, Wenjie Zhang, Minglei Shu, Zewen Kelvin Tuong, Di Yu

{"title":"Feature-preserving manifold approximation and projection to analyze single-cell data","authors":"Yang Yang, Jialei Gong, Hongjian Sun, Amos Choo, Jessica Cara Mar, Yunbo Wei, Yu Zhang, Wenjie Zhang, Minglei Shu, Zewen Kelvin Tuong, Di Yu","doi":"10.1038/s43588-026-00970-6","DOIUrl":"10.1038/s43588-026-00970-6","url":null,"abstract":"Visualizing single-cell data supports understanding cellular heterogeneity and dynamics. Uniform manifold approximation and projection (UMAP) and t-distributed stochastic neighbor embedding (t-SNE) reveal clustering structures but often fail to preserve underlying gene-level information. Here we introduce FeatureMAP (feature-preserving manifold approximation and projection), a framework that enhances manifold learning through pairwise tangent space embedding. By integrating UMAP with principal component analysis, FeatureMAP retains clustering structures and feature variation in a low-dimensional representation. It presents three key analytic concepts: gene contribution, gene variation trajectory, and core versus transition states. Gene contribution and gene variation trajectory are derived by estimating and projecting feature loadings or variation, whereas core and transition states are computationally defined using FeatureMAP’s topological properties, including density, curvature and betweenness centrality. These concepts enable differential gene variation(DGV) analysis that highlights regulatory genes driving transitions between cell states. Demonstrated on synthetic and real single-cell RNA sequencing data from pancreatic development and T cell exhaustion, FeatureMAP improves analyses of trajectories and crucial regulatory genes. FeatureMAP improves single-cell data analysis by preserving gene-level information within the underlying manifold structure. This enables more accurate identification of transitional cell states, differentiation trajectories and key regulatory features across dynamic processes.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 5","pages":"478-496"},"PeriodicalIF":18.3,"publicationDate":"2026-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147719037","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

DualGPT-AB: a dual-stage generative optimization framework for therapeutic antibody design DualGPT-AB：治疗性抗体设计的双阶段生成优化框架。

IF 20.3

Nature computational science Pub Date : 2026-04-15 DOI: 10.1038/s43588-026-00976-0

Dongna Xie, Siyuan Chen, Xi Zeng, Dazhi Lu, Shaoqing Jiao, Shuyuan Xiao, Jiaming Liu, Jianye Hao, Hui Dai, Jiajie Peng

{"title":"DualGPT-AB: a dual-stage generative optimization framework for therapeutic antibody design","authors":"Dongna Xie, Siyuan Chen, Xi Zeng, Dazhi Lu, Shaoqing Jiao, Shuyuan Xiao, Jiaming Liu, Jianye Hao, Hui Dai, Jiajie Peng","doi":"10.1038/s43588-026-00976-0","DOIUrl":"10.1038/s43588-026-00976-0","url":null,"abstract":"Realizing the therapeutic potential of antibodies requires simultaneously optimizing multiple properties, such as antigen-binding specificity, viscosity, clearance and immunogenicity. However, existing methods used for this task are time consuming and resource intensive, often struggling to balance these properties. Here we propose DualGPT-AB, a dual-stage conditional generative pre-trained transformer (GPT) framework for therapeutic antibody design. DualGPT-AB leverages a conditional GPT to model sequence–property relationships by representing the desired properties as learnable embeddings, while incorporating a reinforcement learning strategy to promote antibody sequence exploration and improve optimization efficiency. Computational experiments show that DualGPT-AB is capable of generating antibody heavy chain complementarity-determining region 3 (CDRH3) sequences fulfilling multiple desired properties. Notably, 8 out of 100 randomly selected antibodies from our designed candidate library exhibit excellent HER2-binding affinities. Wet-laboratory validation confirms that DualGPT-AB identifies antibodies with enhanced tumoricidal activity compared with Herceptin, a widely used antibody drug for treating HER2-positive cancers. Overall, DualGPT-AB is a promising approach for advancing artificial intelligence-driven therapeutic antibody development. DualGPT-AB is a dual-stage generative framework for therapeutic antibody design that enables multi-property optimization of CDRH3 sequences. Experiments validate designed antibodies with favorable HER2-binding affinity and tumoricidal activity.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 7","pages":"704-717"},"PeriodicalIF":20.3,"publicationDate":"2026-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147693957","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

Universal restoration of medical images 医学图像的普遍恢复。

IF 18.3

Nature computational science Pub Date : 2026-04-03 DOI: 10.1038/s43588-026-00975-1

Yide Zhang

引用次数: 0

Benchmarking alignment methods for spatial transcriptomics data 空间转录组学数据的基准比对方法。

IF 18.3

Nature computational science Pub Date : 2026-04-03 DOI: 10.1038/s43588-026-00977-z

Yunzhi Yan, Tianyi Gu, Chengcheng Sun, Yinghao Zhang, Yan Cui, Senlin Lin, Qi Zou, Yixuan Du, Chuangyi Han, Kairan Kang, Sijie Li, Yihan Zhao, Zhihong Lin, Zhiyuan Yuan, Bin-Zhi Qian

{"title":"Benchmarking alignment methods for spatial transcriptomics data","authors":"Yunzhi Yan, Tianyi Gu, Chengcheng Sun, Yinghao Zhang, Yan Cui, Senlin Lin, Qi Zou, Yixuan Du, Chuangyi Han, Kairan Kang, Sijie Li, Yihan Zhao, Zhihong Lin, Zhiyuan Yuan, Bin-Zhi Qian","doi":"10.1038/s43588-026-00977-z","DOIUrl":"10.1038/s43588-026-00977-z","url":null,"abstract":"Reconstructing the three-dimensional molecular architecture of tissues from two-dimensional spatial transcriptomics slices is a central goal in spatial biology. Spatial alignment, the computational registration of multiple tissue slices using their spatial coordinates and gene expression profiles, provides the foundational framework for this integrative perspective. Although numerous alignment methods have emerged, a comprehensive benchmark to guide their application has been notably absent. Here we address this by systematically evaluating a diverse suite of leading methods. Executing 295 distinct alignment tasks across diverse datasets and technologies, our framework quantifies method accuracy, efficiency, usability and robustness, while also assessing the downstream impact of alignment quality. Crucially, our study systematically investigates performance in challenging real-world scenarios, uncovering substantial limitations in current tools. To address these bottlenecks, we propose and validate effective mitigation strategies. Finally, we provide practical guidelines to assist researchers in selecting the optimal alignment method and optimizing their analytical workflows. A benchmark to evaluate spatial alignment methods for spatial transcriptomics shows that the best approach is dataset dependent and offers actionable guidance for the future development of methods.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 5","pages":"524-541"},"PeriodicalIF":18.3,"publicationDate":"2026-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147617211","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

To appeal or not to appeal 上诉或不上诉。

IF 18.3

Nature computational science Pub Date : 2026-03-27 DOI: 10.1038/s43588-026-00980-4

引用次数: 0

Scaling and quantization of large-scale foundation model enables resource-efficient predictions in network biology 大规模基础模型的规模化和量化使网络生物学中的资源效率预测成为可能。

IF 18.3

Nature computational science Pub Date : 2026-03-27 DOI: 10.1038/s43588-026-00972-4

Han Chen, Madhavan S. Venkatesh, Javier Gόmez Ortega, Siddharth V. Mahesh, Tarak N. Nandi, Ravi K. Madduri, Karin Pelka, Christina V. Theodoris

{"title":"Scaling and quantization of large-scale foundation model enables resource-efficient predictions in network biology","authors":"Han Chen, Madhavan S. Venkatesh, Javier Gόmez Ortega, Siddharth V. Mahesh, Tarak N. Nandi, Ravi K. Madduri, Karin Pelka, Christina V. Theodoris","doi":"10.1038/s43588-026-00972-4","DOIUrl":"10.1038/s43588-026-00972-4","url":null,"abstract":"Foundation models for network biology are pretrained on large-scale biological data to enable context-aware predictions in a diverse array of downstream tasks through transfer learning. However, increasing model sizes with the expansion of available pretraining data also increases the computational resources required for fine-tuning and inference in downstream applications. Here we first assemble a corpus comprising ~104 million human single-cell transcriptomes from a broad range of tissues and diseases and pretrain successively larger models, defining the scaling laws for transcriptional masked learning. We then demonstrate that model quantization preserves the contextual gene and cell embedding space of the full-precision model, matching performance in zero-shot and fine-tuning applications while requiring only 15% of the time and 34% of the memory as the full model for fine-tuning with the same batch size. Overall, model quantization represents an effective method for resource-efficient fine-tuning and inference while preserving biological knowledge. The authors demonstrate that the accuracy of predictions in network biology scales with larger foundation models pretrained with larger, more diverse data and that quantization enables resource-efficient predictions while preserving biological knowledge.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 5","pages":"450-463"},"PeriodicalIF":18.3,"publicationDate":"2026-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.comhttps://www.nature.com/articles/s43588-026-00972-4.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147535126","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

HorusEye: a self-supervised foundation model for generalizable X-ray tomography restoration HorusEye：一种自我监督的x射线断层扫描复原基础模型。

IF 18.3

Nature computational science Pub Date : 2026-03-27 DOI: 10.1038/s43588-026-00973-3

Yuetan Chu, Longxi Zhou, Gongning Luo, Kai Kang, Suyu Dong, Zhongyi Han, Lianming Wu, Xianglin Meng, Changchun Yang, Xin Guo, Yuan Cheng, Yuan Qi, Xin Liu, Dexuan Xie, Yue Li, Ricardo Henao, Xigang Xiao, Shaodong Cao, Gianluca Setti, Zhaowen Qiu, Xin Gao

{"title":"HorusEye: a self-supervised foundation model for generalizable X-ray tomography restoration","authors":"Yuetan Chu, Longxi Zhou, Gongning Luo, Kai Kang, Suyu Dong, Zhongyi Han, Lianming Wu, Xianglin Meng, Changchun Yang, Xin Guo, Yuan Cheng, Yuan Qi, Xin Liu, Dexuan Xie, Yue Li, Ricardo Henao, Xigang Xiao, Shaodong Cao, Gianluca Setti, Zhaowen Qiu, Xin Gao","doi":"10.1038/s43588-026-00973-3","DOIUrl":"10.1038/s43588-026-00973-3","url":null,"abstract":"X-ray tomography is widely used across scientific and clinical domains, yet image degradation remains a major obstacle to reliable analysis, particularly under low-dose or data-scarce conditions. Existing restoration methods are typically designed for specific modalities and predefined degradation, limiting their generalizability. Here we show that image restoration can instead be formulated as learning realistic, nonparametric acquisition degradation processes directly from data. We introduce HorusEye, a self-supervised foundation model for X-ray tomography restoration that leverages interslice contrastive pretraining to jointly learn structural priors and degradation without paired supervision or predefined assumptions. Trained on over 100 million images, HorusEye generalizes across diverse modalities, restoration tasks and previously unseen imaging modalities, consistently outperforming task-specific approaches. Extensive evaluations demonstrate improved photon efficiency and recovery of high-frequency information. Clinical studies further demonstrate enhanced detectability of low-contrast anatomy and lesions, as well as improved performance on downstream tasks, highlighting HorusEye as a general postprocessing tool for X-ray tomography. HorusEye is a foundation model for universal X-ray tomography restoration that learns realistic degradation directly from data. It supports imaging at substantially lower doses and reduces hardware requirements while improving expert analysis and downstream AI performance.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 4","pages":"372-387"},"PeriodicalIF":18.3,"publicationDate":"2026-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147535097","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