Nature computational science最新文献_第6页

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

Local pseudopotentials accelerate neural network wavefunction simulations 局部伪势加速神经网络波函数模拟。

IF 20.3

Nature computational science Pub Date : 2026-07-13 DOI: 10.1038/s43588-026-01027-4

引用次数: 0

Protein fitness prediction with language models 基于语言模型的蛋白质适应度预测。

IF 20.3

Nature computational science Pub Date : 2026-07-13 DOI: 10.1038/s43588-026-01011-y

Tong Wang

引用次数: 0

Spatial transcriptome for cell segmentation 用于细胞分割的空间转录组。

IF 20.3

Nature computational science Pub Date : 2026-07-10 DOI: 10.1038/s43588-026-01022-9

Nannan Wang (, ), Dayong Jin (, )

引用次数: 0

Empowering neural network-based quantum Monte Carlo with local pseudopotentials 用局部伪势赋能基于神经网络的量子蒙特卡罗。

IF 20.3

Nature computational science Pub Date : 2026-07-10 DOI: 10.1038/s43588-026-01008-7

Weizhong Fu, Ryunosuke Fujimaru, Ruichen Li, Yuzhi Liu, Xuelan Wen, Xiang Li, Kenta Hongo, Liwei Wang, Tom Ichibha, Ryo Maezono, Ji Chen, Weiluo Ren

{"title":"Empowering neural network-based quantum Monte Carlo with local pseudopotentials","authors":"Weizhong Fu, Ryunosuke Fujimaru, Ruichen Li, Yuzhi Liu, Xuelan Wen, Xiang Li, Kenta Hongo, Liwei Wang, Tom Ichibha, Ryo Maezono, Ji Chen, Weiluo Ren","doi":"10.1038/s43588-026-01008-7","DOIUrl":"10.1038/s43588-026-01008-7","url":null,"abstract":"Neural network-based quantum Monte Carlo (NNQMC), an emerging method for solving many-body quantum systems with high accuracy, has been mainly applied to small systems owing to demanding computation requirements. Here we introduce a framework based on local pseudopotentials to break through such limitation, improving the computational efficiency and scalability of NNQMC. The incorporation of local pseudopotentials reduces the number of electrons treated in neural network and also achieves better relative energy accuracy than all electron NNQMC calculations for complex systems. This counterintuitive outcome is made possible by the distinctive characteristics inherent to NNQMC. Notably, by avoiding costly integration terms, this approach is also substantially more efficient than its widely used semilocal counterparts. Our approach enables the reliable treatment of large and challenging systems, such as the $${text{Fe}}_{4}{text{S}}_{4}{({text{SCH}}_{3})}_{4}$$ iron–sulfur cluster. Overall, our findings demonstrate that the synergy between NNQMC and local pseudopotentials substantially expands the scope of accurate ab initio calculations. This study shows that local pseudopotentials boost the efficiency of neural network-based quantum Monte Carlo while preserving accuracy, enabling reliable simulations of large, challenging iron–sulfur clusters.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 9","pages":"1005-1014"},"PeriodicalIF":20.3,"publicationDate":"2026-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148427184","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

Integrating cytological images and spatial transcriptomics for cell segmentation with DISSECT 整合细胞学图像和空间转录组学的细胞分割与解剖。

IF 20.3

Nature computational science Pub Date : 2026-07-09 DOI: 10.1038/s43588-026-01020-x

Yufeng He, Yanping Zhao, Rui Zhang, Heli Yang, Zongxu Zhang, Pengfei Ren, Ce Luo, Peng Zhang, Zhe Zhang, Sen Hou, Zhaode Bu, Yuan Luo, Deng Pan, Zexian Zeng

{"title":"Integrating cytological images and spatial transcriptomics for cell segmentation with DISSECT","authors":"Yufeng He, Yanping Zhao, Rui Zhang, Heli Yang, Zongxu Zhang, Pengfei Ren, Ce Luo, Peng Zhang, Zhe Zhang, Sen Hou, Zhaode Bu, Yuan Luo, Deng Pan, Zexian Zeng","doi":"10.1038/s43588-026-01020-x","DOIUrl":"10.1038/s43588-026-01020-x","url":null,"abstract":"Advances in imaging- and sequencing-based spatial transcriptomics have increased molecular throughput and resolution, enabling the measurement and analysis of spatial transcriptomes at single-cell resolution. However, accurate cell segmentation remains challenging because cell morphology, tissue processing and staining methods vary across samples and platforms, limiting the accuracy and generalizability of existing algorithms. Here we show that DISSECT, a cell segmentation model integrating cytological images with spatial transcriptomic profiles, improves spatial single-cell transcriptome reconstruction. DISSECT uses a pretrained deep generative model to denoise multiscale image features, predicts cell instances with an instance-aware detection module and applies image- and transcriptome-derived gradient fields to refine segmentation masks. Benchmarking across multiple datasets showed that DISSECT achieved higher mean average precision than several existing segmentation tools. We further applied DISSECT to three pairs of gastric adenocarcinoma samples collected before and after anti-PD-1 treatment and profiled by Stereo-seq, illustrating its utility for downstream spatial biological interpretation. The authors propose a framework that integrates cytological imaging and spatial transcriptomics to improve cell segmentation and enable reconstruction of spatial single-cell transcriptomes.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 7","pages":"718-731"},"PeriodicalIF":20.3,"publicationDate":"2026-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148427101","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

NEP89: universal neuroevolution potential for inorganic and organic materials across 89 elements NEP89：跨89种元素的无机和有机材料的普遍神经进化潜力。

IF 20.3

Nature computational science Pub Date : 2026-07-08 DOI: 10.1038/s43588-026-01009-6

Ting Liang, Ke Xu, Eric Lindgren, Zherui Chen, Rui Zhao, Jiahui Liu, Esmée Berger, Benrui Tang, Bohan Zhang, Yanzhou Wang, Keke Song, Penghua Ying, Nan Xu, Haikuan Dong, Shunda Chen, Paul Erhart, Zheyong Fan, Tapio Ala-Nissila, Jianbin Xu

{"title":"NEP89: universal neuroevolution potential for inorganic and organic materials across 89 elements","authors":"Ting Liang, Ke Xu, Eric Lindgren, Zherui Chen, Rui Zhao, Jiahui Liu, Esmée Berger, Benrui Tang, Bohan Zhang, Yanzhou Wang, Keke Song, Penghua Ying, Nan Xu, Haikuan Dong, Shunda Chen, Paul Erhart, Zheyong Fan, Tapio Ala-Nissila, Jianbin Xu","doi":"10.1038/s43588-026-01009-6","DOIUrl":"10.1038/s43588-026-01009-6","url":null,"abstract":"While machine-learned interatomic potentials offer near-quantum-mechanical accuracy for atomistic simulations, many are material-specific or computationally intensive, limiting their broader use. Here we introduce NEP89, a foundation model based on neuroevolution potential architecture, delivering near-empirical-potential speed and high accuracy across 89 elements. A compact yet comprehensive training dataset covering inorganic and organic materials was curated through descriptor-space subsampling and iterative refinement across multiple datasets. NEP89 achieves competitive accuracy compared with representative foundation models while being three to four orders of magnitude more computationally efficient, enabling previously impractical large-scale atomistic simulations of inorganic and organic systems. In addition to its out-of-the-box applicability to diverse scenarios, including million-atom-scale compression of compositionally complex alloys, ion diffusion in solid-state electrolytes and water, rocksalt dissolution, methane combustion and protein–ligand dynamics, NEP89 also supports fine-tuning for rapid adaptation to user-specific applications, such as mechanical, thermal, structural and spectral properties of two-dimensional materials, metallic glasses and organic crystals. The researchers developed NEP89, a foundation model for large-scale molecular dynamics across 89 elements with near-quantum-level accuracy and high efficiency. It can be rapidly adapted to address complex challenges in inorganic and organic materials.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 7","pages":"789-801"},"PeriodicalIF":20.3,"publicationDate":"2026-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148414139","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

Ultrafast and ultralarge distance-based phylogenetics using DIPPER 使用DIPPER进行超快和超大距离的系统发育。

IF 20.3

Nature computational science Pub Date : 2026-07-08 DOI: 10.1038/s43588-026-01015-8

Sumit Walia, Zexing Chen, Yu-Hsiang Tseng, Yatish Turakhia

{"title":"Ultrafast and ultralarge distance-based phylogenetics using DIPPER","authors":"Sumit Walia, Zexing Chen, Yu-Hsiang Tseng, Yatish Turakhia","doi":"10.1038/s43588-026-01015-8","DOIUrl":"10.1038/s43588-026-01015-8","url":null,"abstract":"Distance-based methods are commonly used to reconstruct phylogenies for various applications owing to their excellent speed, scalability and theoretical guarantees. However, classical de novo algorithms are hindered by cubic time and quadratic memory complexity, making them impractical for emerging datasets containing millions of sequences. Here we present DIPPER, a distance-based tool for phylogenetic reconstruction on graphics processing units (GPUs), designed to maintain high accuracy and a low memory footprint. DIPPER employs a divide-and-conquer strategy, a placement strategy and an on-the-fly distance calculator that improve runtime and memory complexity to O(N.log(N)) and O(N), respectively, with N taxa. DIPPER also maintains a low memory footprint on the GPU that is independent of the number of taxa. DIPPER outperforms existing methods in speed and memory efficiency on both simulated and real-world datasets, enabling the reconstruction of 10 million sequences in under 7 h on a single NVIDIA RTX A6000 GPU. This study introduces DIPPER, a GPU-accelerated tool that efficiently reconstructs evolutionary trees from tens of millions of genome sequences, extending scalability by one to two orders of magnitude beyond that of existing tools.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 8","pages":"816-825"},"PeriodicalIF":20.3,"publicationDate":"2026-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148414086","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

Computing inspired by the brain: a journey from algorithms to organoids 受大脑启发的计算：从算法到类器官的旅程。

IF 20.3

Nature computational science Pub Date : 2026-07-03 DOI: 10.1038/s43588-026-01012-x

Paris Brown, Shyni Varghese

{"title":"Computing inspired by the brain: a journey from algorithms to organoids","authors":"Paris Brown, Shyni Varghese","doi":"10.1038/s43588-026-01012-x","DOIUrl":"10.1038/s43588-026-01012-x","url":null,"abstract":"The human brain has long served as a blueprint for computation, guiding evolution from early symbolic systems to modern deep learning models. Despite these advances, traditional computing systems remain fundamentally limited in mirroring the remarkable flexibility, parallel processing and energy efficiency of the human brain. To address these limitations, neuromorphic computing was developed, which mimics the architecture and signaling behavior of biological neurons. Building on this foundation, a new frontier is now emerging—organoid intelligence (OI). OI uses lab-grown brain cellular structures, such as living neural organoids with electrical activity, synapse formation and primitive learning, as a substrate for computation. Here we trace the evolution of brain-inspired computing from symbolic logic systems to artificial neural networks, neuromorphic processors and finally biohybrid computers that incorporate living neural structures. We explore the transformative potential of OI along with the substantial technical, biological and ethical challenges it presents. Organoid intelligence represents a new frontier for computing inspired by the human brain. This Review discusses the evolution of brain-inspired computing along with potential challenges for future development.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 7","pages":"686-696"},"PeriodicalIF":20.3,"publicationDate":"2026-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148383307","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

Systematic bottom-up coarse-graining of hydrated excess proton transport across scales 系统的自下而上的水合过量质子跨尺度输运粗粒化。

IF 20.3

Nature computational science Pub Date : 2026-07-01 DOI: 10.1038/s43588-026-01006-9

Jaehyeok Jin, Zhefu Li, Gregory A. Voth

{"title":"Systematic bottom-up coarse-graining of hydrated excess proton transport across scales","authors":"Jaehyeok Jin, Zhefu Li, Gregory A. Voth","doi":"10.1038/s43588-026-01006-9","DOIUrl":"10.1038/s43588-026-01006-9","url":null,"abstract":"Proton transport is a multiscale phenomenon critical to chemistry, physics, biology and materials science, yet development of a high-fidelity multiscale model bridging quantum to mesoscale dynamics remains a major challenge owing to reactive bonding rearrangements in the Grotthuss proton shuttling. Here we present the Hydronium Ultra-coarse-grained Model with Improved Dynamics (HUMID), which enables the investigation of hydrated protons across much larger spatiotemporal scales via bottom-up coarse-graining (CG) that recapitulates structure and dynamics at a reduced resolution. The unique structural correlations and dynamics arising from the hydronium cation are captured by introducing internal states to the CG sites, as derived from quantum mechanics, with a systematic design principle for determining these internal states. HUMID accurately reproduces diffusion and hydronium time correlation functions with a three-order-of-magnitude computational speedup. This model enables the study of mesoscale phenomena governed by proton transport and establishes design principles for reactive CG models. HUMID is a bottom-up coarse-grained framework for proton transport that bridges the quantum and molecular scales, enabling simulations of reactive transport dynamics across diverse environments with over 1,000-fold greater computational efficiency.","PeriodicalId":74246,"journal":{"name":"Nature computational science","volume":"6 7","pages":"767-777"},"PeriodicalIF":20.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148371296","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