Nature Machine Intelligence最新文献_第2页

Towards unveiling sensitive and decisive patterns in explainable AI with a case study in geometric deep learning

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-17 DOI: 10.1038/s42256-025-00998-9

Jiajun Zhu, Siqi Miao, Rex Ying, Pan Li

{"title":"Towards unveiling sensitive and decisive patterns in explainable AI with a case study in geometric deep learning","authors":"Jiajun Zhu, Siqi Miao, Rex Ying, Pan Li","doi":"10.1038/s42256-025-00998-9","DOIUrl":"10.1038/s42256-025-00998-9","url":null,"abstract":"The interpretability of machine learning models has gained increasing attention, particularly in scientific domains where high precision and accountability are crucial. This research focuses on distinguishing between two critical data patterns—sensitive patterns (model related) and decisive patterns (task related)—which are commonly used as model interpretations but often lead to confusion. Specifically, this study compares the effectiveness of two main streams of interpretation methods: post-hoc methods and self-interpretable methods, in detecting these patterns. Recently, geometric deep learning (GDL) has shown superior predictive performance in various scientific applications, creating an urgent need for principled interpretation methods. Here, therefore, we conduct our study using several representative GDL applications as case studies. We evaluate 13 interpretation methods applied to 3 major GDL backbone models, using 4 scientific datasets to assess how well these methods identify sensitive and decisive patterns. Our findings indicate that post-hoc methods tend to provide interpretations better aligned with sensitive patterns, whereas certain self-interpretable methods exhibit strong and stable performance in detecting decisive patterns. Moreover, our study offers valuable insights into improving the reliability of these interpretation methods. For example, ensembling post-hoc interpretations from multiple models trained on the same task can effectively uncover the task’s decisive patterns. Interpreting decisions made by machine learning systems remains difficult. Here Zhu et al. test interpretability methods on their ability to identify model-related and task-related patterns.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 3","pages":"471-483"},"PeriodicalIF":18.8,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143635717","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Synergy-based robotic quadruped leveraging passivity for natural intelligence and behavioural diversity

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-17 DOI: 10.1038/s42256-025-00988-x

Francesco Stella, Mickaël M. Achkar, Cosimo Della Santina, Josie Hughes

{"title":"Synergy-based robotic quadruped leveraging passivity for natural intelligence and behavioural diversity","authors":"Francesco Stella, Mickaël M. Achkar, Cosimo Della Santina, Josie Hughes","doi":"10.1038/s42256-025-00988-x","DOIUrl":"10.1038/s42256-025-00988-x","url":null,"abstract":"Quadrupedal animals show remarkable capabilities in traversing diverse terrains and display a range of behaviours and gait patterns. Achieving similar performance by exploiting the natural dynamics of the system is a key goal for robotics researchers. Here we show a bioinspired approach to the design of quadrupeds that seeks to exploit the body and the passive properties of the robot while maintaining active controllability on the system through minimal actuation. Utilizing an end-to-end computational design pipeline, neuromechanical couplings recorded in biological quadrupeds are translated into motor synergies, allowing minimal actuation to control the full structure via multijoint compliant mechanical couplings. Using this approach, we develop PAWS, a passive automata with synergies. By leveraging the principles of motor synergies, the design incorporates variable stiffness, anatomical insights and self-organization to simplify control while maximizing its capabilities. The resulting synergy-based quadruped requires only four actuators and exhibits emergent, animal-like dynamical responses, including passive robustness to environmental perturbations and a wide range of actuated behaviours. The finding contributes to the development of machine physical intelligence and provides robots with more efficient and natural-looking robotic locomotion by combining synergistic actuation, compliant body properties and embodied compensatory strategies. Stella, Achkar and colleagues present a bio-inspired quadruped robot that uses passive dynamics, motor synergies, variable stiffness and self-organization to achieve robust, adaptive, animal-like movement with just four actuators.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 3","pages":"386-399"},"PeriodicalIF":18.8,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.com/articles/s42256-025-00988-x.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143635721","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A comprehensive large-scale biomedical knowledge graph for AI-powered data-driven biomedical research 用于人工智能驱动的数据驱动生物医学研究的综合大规模生物医学知识图谱

IF 23.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-17 DOI: 10.1038/s42256-025-01014-w

Yuan Zhang, Xin Sui, Feng Pan, Kaixian Yu, Keqiao Li, Shubo Tian, Arslan Erdengasileng, Qing Han, Wanjing Wang, Jianan Wang, Jian Wang, Donghu Sun, Henry Chung, Jun Zhou, Eric Zhou, Ben Lee, Peili Zhang, Xing Qiu, Tingting Zhao, Jinfeng Zhang

{"title":"A comprehensive large-scale biomedical knowledge graph for AI-powered data-driven biomedical research","authors":"Yuan Zhang, Xin Sui, Feng Pan, Kaixian Yu, Keqiao Li, Shubo Tian, Arslan Erdengasileng, Qing Han, Wanjing Wang, Jianan Wang, Jian Wang, Donghu Sun, Henry Chung, Jun Zhou, Eric Zhou, Ben Lee, Peili Zhang, Xing Qiu, Tingting Zhao, Jinfeng Zhang","doi":"10.1038/s42256-025-01014-w","DOIUrl":"https://doi.org/10.1038/s42256-025-01014-w","url":null,"abstract":"<p>To address the rapid growth of scientific publications and data in biomedical research, knowledge graphs (KGs) have become a critical tool for integrating large volumes of heterogeneous data to enable efficient information retrieval and automated knowledge discovery. However, transforming unstructured scientific literature into KGs remains a significant challenge, with previous methods unable to achieve human-level accuracy. Here we used an information extraction pipeline that won first place in the LitCoin Natural Language Processing Challenge (2022) to construct a large-scale KG named iKraph using all PubMed abstracts. The extracted information matches human expert annotations and significantly exceeds the content of manually curated public databases. To enhance the KG’s comprehensiveness, we integrated relation data from 40 public databases and relation information inferred from high-throughput genomics data. This KG facilitates rigorous performance evaluation of automated knowledge discovery, which was infeasible in previous studies. We designed an interpretable, probabilistic-based inference method to identify indirect causal relations and applied it to real-time COVID-19 drug repurposing from March 2020 to May 2023. Our method identified around 1,200 candidate drugs in the first 4 months, with one-third of those discovered in the first 2 months later supported by clinical trials or PubMed publications. These outcomes are very challenging to attain through alternative approaches that lack a thorough understanding of the existing literature. A cloud-based platform (https://biokde.insilicom.com) was developed for academic users to access this rich structured data and associated tools.</p>","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"28 1","pages":""},"PeriodicalIF":23.8,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143635718","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

From data chaos to precision medicine

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-13 DOI: 10.1038/s42256-025-01015-9

Alexander Schönhuth

引用次数: 0

Transformers and genome language models

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-13 DOI: 10.1038/s42256-025-01007-9

Micaela E. Consens, Cameron Dufault, Michael Wainberg, Duncan Forster, Mehran Karimzadeh, Hani Goodarzi, Fabian J. Theis, Alan Moses, Bo Wang

{"title":"Transformers and genome language models","authors":"Micaela E. Consens, Cameron Dufault, Michael Wainberg, Duncan Forster, Mehran Karimzadeh, Hani Goodarzi, Fabian J. Theis, Alan Moses, Bo Wang","doi":"10.1038/s42256-025-01007-9","DOIUrl":"10.1038/s42256-025-01007-9","url":null,"abstract":"Large language models based on the transformer deep learning architecture have revolutionized natural language processing. Motivated by the analogy between human language and the genome’s biological code, researchers have begun to develop genome language models (gLMs) based on transformers and related architectures. This Review explores the use of transformers and language models in genomics. We survey open questions in genomics amenable to the use of gLMs, and motivate the use of gLMs and the transformer architecture for these problems. We discuss the potential of gLMs for modelling the genome using unsupervised pretraining tasks, specifically focusing on the power of zero- and few-shot learning. We explore the strengths and limitations of the transformer architecture, as well as the strengths and limitations of current gLMs more broadly. Additionally, we contemplate the future of genomic modelling beyond the transformer architecture, based on current trends in research. This Review serves as a guide for computational biologists and computer scientists interested in transformers and language models for genomic data. Micaela Consens et al. discuss and review the recent rise of transformer-based and large language models in genomics. They also highlight promising directions for genome language models beyond the transformer architecture.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 3","pages":"346-362"},"PeriodicalIF":18.8,"publicationDate":"2025-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143608413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Fast, scale-adaptive and uncertainty-aware downscaling of Earth system model fields with generative machine learning

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-13 DOI: 10.1038/s42256-025-00980-5

Philipp Hess, Michael Aich, Baoxiang Pan, Niklas Boers

{"title":"Fast, scale-adaptive and uncertainty-aware downscaling of Earth system model fields with generative machine learning","authors":"Philipp Hess, Michael Aich, Baoxiang Pan, Niklas Boers","doi":"10.1038/s42256-025-00980-5","DOIUrl":"10.1038/s42256-025-00980-5","url":null,"abstract":"Accurate and high-resolution Earth system model (ESM) simulations are essential to assess the ecological and socioeconomic impacts of anthropogenic climate change, but are computationally too expensive to be run at sufficiently high spatial resolution. Recent machine learning approaches have shown promising results in downscaling ESM simulations, outperforming state-of-the-art statistical approaches. However, existing methods require computationally costly retraining for each ESM and extrapolate poorly to climates unseen during training. We address these shortcomings by learning a consistency model that efficiently and accurately downscales arbitrary ESM simulations without retraining in a zero-shot manner. Our approach yields probabilistic downscaled fields at a resolution only limited by the observational reference data. We show that the consistency model outperforms state-of-the-art diffusion models at a fraction of the computational cost and maintains high controllability on the downscaling task. Further, our method generalizes to climate states unseen during training without explicitly formulated physical constraints. A generative machine learning approach is proposed to improve the resolution of Earth system models in an efficient, adaptive and uncertainty-aware manner.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 3","pages":"363-373"},"PeriodicalIF":18.8,"publicationDate":"2025-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.com/articles/s42256-025-00980-5.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143608414","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Data-driven federated learning in drug discovery with knowledge distillation

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-05 DOI: 10.1038/s42256-025-00991-2

Thierry Hanser, Ernst Ahlberg, Alexander Amberg, Lennart T. Anger, Chris Barber, Richard J. Brennan, Alessandro Brigo, Annie Delaunois, Susanne Glowienke, Nigel Greene, Laura Johnston, Daniel Kuhn, Lara Kuhnke, Jean-François Marchaland, Wolfgang Muster, Jeffrey Plante, Friedrich Rippmann, Yogesh Sabnis, Friedemann Schmidt, Ruud van Deursen, Stéphane Werner, Angela White, Joerg Wichard, Tomoya Yukawa

{"title":"Data-driven federated learning in drug discovery with knowledge distillation","authors":"Thierry Hanser, Ernst Ahlberg, Alexander Amberg, Lennart T. Anger, Chris Barber, Richard J. Brennan, Alessandro Brigo, Annie Delaunois, Susanne Glowienke, Nigel Greene, Laura Johnston, Daniel Kuhn, Lara Kuhnke, Jean-François Marchaland, Wolfgang Muster, Jeffrey Plante, Friedrich Rippmann, Yogesh Sabnis, Friedemann Schmidt, Ruud van Deursen, Stéphane Werner, Angela White, Joerg Wichard, Tomoya Yukawa","doi":"10.1038/s42256-025-00991-2","DOIUrl":"10.1038/s42256-025-00991-2","url":null,"abstract":"A main challenge for artificial intelligence in scientific research is ensuring access to sufficient, high-quality data for the development of impactful models. Despite the abundance of public data, the most valuable knowledge often remains embedded within confidential corporate data silos. Although industries are increasingly open to sharing non-competitive insights, such collaboration is often constrained by the confidentiality of the underlying data. Federated learning makes it possible to share knowledge without compromising data privacy, but it has notable limitations. Here, we introduce FLuID (federated learning using information distillation), a data-centric application of federated distillation tailored to drug discovery aiming to preserve data privacy. We validate FLuID in two experiments, first involving public data simulating a virtual consortium and second in a real-world research collaboration between eight pharmaceutical companies. Although the alignment of the models with the partner specific domain remains challenging, the data-driven nature of FLuID offers several avenues to mitigate domain shift. FLuID fosters knowledge sharing among pharmaceutical organizations, paving the way for a new generation of models with enhanced performance and an expanded applicability domain in biological activity predictions. FLuID enables privacy-preserving knowledge sharing in drug discovery using knowledge distillation. The results show that the approach expands applicability domains and fosters collaboration across organizations without compromising data privacy or security.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 3","pages":"423-436"},"PeriodicalIF":18.8,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143546524","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Bridging the gap between machine confidence and human perceptions

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-03-03 DOI: 10.1038/s42256-025-01013-x

Ming Yin

引用次数: 0

A unified deep framework for peptide–major histocompatibility complex–T cell receptor binding prediction

IF 23.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-02-26 DOI: 10.1038/s42256-025-01002-0

Yunxiang Zhao, Jijun Yu, Yixin Su, You Shu, Enhao Ma, Jing Wang, Shuyang Jiang, Congwen Wei, Dongsheng Li, Zhen Huang, Gong Cheng, Hongguang Ren, Jiannan Feng

{"title":"A unified deep framework for peptide–major histocompatibility complex–T cell receptor binding prediction","authors":"Yunxiang Zhao, Jijun Yu, Yixin Su, You Shu, Enhao Ma, Jing Wang, Shuyang Jiang, Congwen Wei, Dongsheng Li, Zhen Huang, Gong Cheng, Hongguang Ren, Jiannan Feng","doi":"10.1038/s42256-025-01002-0","DOIUrl":"https://doi.org/10.1038/s42256-025-01002-0","url":null,"abstract":"<p>Antigen peptides that are presented by a major histocompatibility complex (MHC) and recognized by a T cell receptor (TCR) have an essential role in immunotherapy. Although substantial progress has been made in predicting MHC presentation, accurately predicting the binding interactions between antigen peptides, MHCs and TCRs remains a major computational challenge. In this paper, we propose a unified deep framework (called UniPMT) for peptide, MHC and TCR binding prediction to predict the binding between the peptide and the CDR3 of TCR β in general, presented by class I MHCs. UniPMT is comprehensively validated by a series of experiments and achieved state-of-the-art performance in the peptide–MHC–TCR, peptide–MHC and peptide–TCR binding prediction tasks with up to 15% improvements in area under the precision–recall curve taking the peptide–MHC–TCR binding prediction task as an example. In practical applications, UniPMT shows strong predictive power, correlates well with T cell clonal expansion and outperforms existing methods in neoantigen-specific binding prediction with up to 17.62% improvements in area under the precision–recall curve on experimentally validated datasets. Moreover, UniPMT provides interpretable insights into the identification of key binding sites and the quantification of peptide–MHC–TCR binding probabilities. In summary, UniPMT shows great potential to serve as a useful tool for antigen peptide discovery, disease immunotherapy and neoantigen vaccine design.</p>","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"51 1","pages":""},"PeriodicalIF":23.8,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143495337","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Large language models for scientific discovery in molecular property prediction

IF 18.8 1区计算机科学

Nature Machine Intelligence Pub Date : 2025-02-25 DOI: 10.1038/s42256-025-00994-z

Yizhen Zheng, Huan Yee Koh, Jiaxin Ju, Anh T. N. Nguyen, Lauren T. May, Geoffrey I. Webb, Shirui Pan

{"title":"Large language models for scientific discovery in molecular property prediction","authors":"Yizhen Zheng, Huan Yee Koh, Jiaxin Ju, Anh T. N. Nguyen, Lauren T. May, Geoffrey I. Webb, Shirui Pan","doi":"10.1038/s42256-025-00994-z","DOIUrl":"10.1038/s42256-025-00994-z","url":null,"abstract":"Large language models (LLMs) are a form of artificial intelligence system encapsulating vast knowledge in the form of natural language. These systems are adept at numerous complex tasks including creative writing, storytelling, translation, question-answering, summarization and computer code generation. Although LLMs have seen initial applications in natural sciences, their potential for driving scientific discovery remains largely unexplored. In this work, we introduce LLM4SD, a framework designed to harness LLMs for driving scientific discovery in molecular property prediction by synthesizing knowledge from literature and inferring knowledge from scientific data. LLMs synthesize knowledge by extracting established information from scientific literature, such as molecular weight being key to predicting solubility. For inference, LLMs identify patterns in molecular data, particularly in Simplified Molecular Input Line Entry System-encoded structures, such as halogen-containing molecules being more likely to cross the blood–brain barrier. This information is presented as interpretable knowledge, enabling the transformation of molecules into feature vectors. By using these features with interpretable models such as random forest, LLM4SD can outperform the current state of the art across a range of benchmark tasks for predicting molecular properties. We foresee it providing interpretable and potentially new insights, aiding scientific discovery in molecular property prediction. Zheng et al. developed LLM4SD, a framework using large language models to predict molecular properties. The method leverages the ability of large language models to synthesize knowledge from literature and to reason about scientific data with domain expertise.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 3","pages":"437-447"},"PeriodicalIF":18.8,"publicationDate":"2025-02-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143486029","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0