Nature Computational Science

🚨Our October issue is now live, including an approach for modulating complex oscillations, a Review on using LLMs as human proxies, a framework for predicting brain morphogenesis, and much more! Check it out! www.nature.com/natcomputsci... 📰Cover: www.nature.com/articles/s43...

nature.comMachine learning dataset and benchmark for accurate T cell receptor–pHLA binding prediction - Nature Computational ScienceHi-TPH is a large-scale hierarchical dataset that enables more accurate prediction of T cell receptor binding to antigens using protein language models and reveals how molecular components jointly con...

📢Fuli Feng and colleagues introduce a large-scale hierarchical dataset that enables more accurate prediction of T cell receptor binding to antigens using PLMs, revealing how molecular components jointly contribute to prediction. www.nature.com/articles/s43... #immunosky 🔓 rdcu.be/ZQ6reLwys86a

nature.comMaking the case for a non-primary article - Nature Computational ScienceAuthor-proposed non-primary articles can bring fresh and timely perspectives to the journal. Here, we discuss the factors that editors consider when assessing proposals for these contributions.

📢In our latest Editorial, we discuss how to submit proposals for non-primary submission. www.nature.com/articles/s43... #SciencePublishing #ResearchPublishing

nature.comStructure-coefficient theorem for evolutionary games on dynamic networks - Nature Computational ScienceThe authors establish a structure-coefficient theorem, showing that three universal coefficients capture evolutionary outcomes in changing networks, and identify dynamic core–periphery networks that p...

📢A new work presents a structure-coefficient theorem for evolutionary games on dynamic networks and identifies dynamic core-periphery networks that promote cooperation, altruism, fairness and trust. www.nature.com/articles/s43... #evosky #evobio 🔓 rdcu.be/YmXkfchoZHua

nature.comDynamic networks can promote pro-sociality - Nature Computational ScienceStructure coefficients provide a unified way to identify dynamic network structures that promote cooperation, altruism, fairness and trust.

An accompanying News & Views for this manuscript by @esocrats.bsky.social is also available! www.nature.com/articles/s43... @fwovlaanderen.bsky.social @faribrussels.bsky.social #evosky #evobio 🔓 rdcu.be/5nvco86QkX6a

nature.comHDOCK-Multimer: improved structure prediction of large protein complexes by integrating docking and combinatorial assembly - Nature Computational ScienceThis study presents HDOCK-Multimer, a general framework for structure prediction of large protein complexes by integrating ab initio docking and combinatorial assembly, substantially outperforming sta...

📢Sheng-You Huang and colleagues present HDOCK-Multimer, a general framework for structure prediction of large protein complexes by integrating ab initio docking and combinatorial assembly. www.nature.com/articles/s43... 🔓

nature.comHIP: Hessian Interatomic Potentials without derivatives - Nature Computational ScienceThe authors introduce an extension of machine learning interatomic potentials that predicts molecular Hessians directly, improving accuracy and speed compared in transition-state searches, geometry op...

📢Out now! @andreasburger.bsky.social, @thematterlab.bsky.social and colleagues develop a direct approach for accurately and efficiently predicting molecular Hessians. www.nature.com/articles/s43... 🔓 #chemsky

nature.comLess can be more for predicting properties with large language models - Nature Computational ScienceResearchers show that large language models cannot learn from spatial coordinates regardless of model or data scale. Specialized geometric neural networks consistently outperform language models for m...

📢Out now! Kevin Maik Jablonka and colleagues identify fundamental limitations in LLM's ability to learn from spatial data across models and data scale. @uni-jena.de www.nature.com/articles/s43... 🔓

nature.comEpidemiological modeling in the age of artificial intelligence - Nature Computational ScienceArtificial intelligence is enabling epidemic models to learn across outbreaks. The next challenge is ensuring that these models learn and transfer not mere statistical patterns, but epidemiological me...

📢Can AI learn from past epidemics to better understand future outbreaks? Max Lau et al. contribute to our 5-year Series and explore how epidemic models can move beyond prediction to capture the mechanisms driving disease spread. www.nature.com/articles/s43... #episky 🔓 rdcu.be/qY1bFJ1jEhna

nature.comAdvancing chemical reaction prediction in data-scarce drug discovery with active and geometric deep learning - Nature Computational ScienceResearchers pair active learning with geometric deep learning to predict the outcome and site selectivity of chemical reactions from scarce data, guiding experiments that accelerate drug discovery and...

📢Out now! Gisbert Schneider and colleagues pair active learning and geometric deep learning to predict the outcome and site selectivity of chemical reactions from scarce data, helping to accelerate the drug discovery process. www.nature.com/articles/s43... 🔓 #chemsky

nature.comCellART: a unified framework for extracting single-cell information from high-resolution spatial transcriptomics - Nature Computational ScienceCellART is a deep learning framework that recovers individual cells from high-resolution spatial transcriptomics data, predicting cell boundaries and cell types together across different platforms and...

📢Can Yang et al. present CellART, a DL framework that recovers individual cells from high-resolution spatial transcriptomics data, predicting cell boundaries and cell types together across different platforms and tissue types. www.nature.com/articles/s43... 🖥️ 🧬 🔓 rdcu.be/kCCcn3nZRdBa

nature.comPhotonic neuromorphic learning via generalized in situ physical gradient descent - Nature Computational ScienceThis study reports a physics-based in situ learning method that trains artificial neural networks directly within the photonic integrated circuits, turning diverse optical circuits into efficient AI l...

📢Lu Fang and colleagues develop a physics-based in situ learning method that trains ANNs directly within the photonic integrated circuits, turning diverse optical circuits into efficient AI learning hardware. www.nature.com/articles/s43... ⚛️ 🔓 rdcu.be/FiSaE96yYCsa

nature.comModulatability of complex oscillators - Nature Computational ScienceThis Article reports a property-based framework with a rigorous mathematical foundation for modulating complex rhythms. It identifies the fewest parameters for producing desired spatiotemporal behavio...

📢 @boweiqin.bsky.social, @chaosfudan.bsky.social and colleagues present a framework that identifies modulation parameters based on desired spatiotemporal properties. www.nature.com/articles/s43... @iics-fudan.bsky.social #complexity ⚛️ 🔓 rdcu.be/AbW2zo3o1Uya

22 Sept

🚨Our September issue is now live, including a single-cell multimodal data integration tool, a foundation model for pathology image segmentation, a study on self-rankings for peer review, and much more! Check it out! www.nature.com/natcomputsci... 📰Cover: www.nature.com/articles/s43...

nature.comComputational active materials and embodied intelligence in extreme conditions - Nature Computational ScienceEmbodied intelligence has enabled machines to act autonomously, yet intelligence remains largely confined to digital models, leaving robotic bodies as passive executors. In extreme environments, such ...

22 Sept

📢In our latest 5-year Series Comment, Yifan Yang and Fan Xu discuss how the computational design of active materials can expand robotic capabilities in extreme environments, such as the human brain. www.nature.com/articles/s43... #MaterialsScience

nature.comReproducibility in the era of large language models - Nature Computational ScienceReproducibility has long been a concern in machine learning research. The rise of large language models adds new layers of complexity, underscoring the need for clearer reporting standards and communi...

22 Sept

📢In our latest Editorial, we discuss how LLMs have made reproducibility more challenging, as well as provide some recommendations on how to report LLM-based research results. www.nature.com/articles/s43... #ArtificialIntelligence #ResearchPublishing

nature.comAlgorithmic systems, human agency and the future of platform research - Nature Computational ScienceWhile there are widespread concerns about the potential negative societal impacts of algorithm-driven systems, a more balanced approach to understanding online content consumption and engagement that ...

21 Sept

📢 @homahm.bsky.social, @duncanjwatts.bsky.social and colleagues review the literature on the role of algorithmic recommendation and show that evidence linking algorithmic curation to problematic outcomes is inconclusive, calling for the need for more research. www.nature.com/articles/s43... #cssky

nature.comLarge language models as human proxies - Nature Computational ScienceLarge language models can potentially serve as human proxies across various roles, enabling nuanced simulations of behavior. This Review categorizes their applications, stressing distinct evaluation m...

21 Sept

📢Out now! Nikita Karetnikov, Iyad Rahwan, and Davor Svetinovic discuss how LLMs are reshaping our understanding of human behavior by acting as proxies in various roles. www.nature.com/articles/s43... @wuvienna.bsky.social #cssky 🔓 rdcu.be/YZzP5Dn6Qt4a

nature.comRevisiting molecular descriptors with TDiMS for interpretable intramolecular interactions based on substructure pairs - Nature Computational ScienceA descriptor, TDiMS, is developed to summarize the enumerated pairwise topological distances between molecular substructures and identify interpretable molecular features for materials discovery.

21 Sept

📢Lisa Hamada and colleagues present TDiMS, a molecular descriptor that summarizes the enumerated pairwise topological distances between molecular substructures, capturing nonlocal interactions. www.nature.com/articles/s43... 🔓 #chemsky

nature.comSegment anything in pathology images with natural language - Nature Computational ScienceThis Resource introduces PathSegmentor—a foundation model that segments 160 pathological categories via natural language prompts, generalizes across datasets and clinical cohorts, and enables scalable...

10 Sept

📢A new Resource introduces PathSegmentor, a foundation model for segmenting structures across different anatomical regions and spatial scales. www.nature.com/articles/s43... #Bioimaging 🔓 rdcu.be/ttPkGmX7yy6a

nature.comWhen pathology segmentation learns to listen - Nature Computational ScienceA natural-language-guided pathology segmentation model is developed to link pathological language with pathology image content to produce semantic masks, offering a path toward computational pathology...

10 Sept

An accompanying News & Views by Wei Shen for this paper is also available! www.nature.com/articles/s43... #Bioimaging 🔓 rdcu.be/kKxBXwwrKzBa

nature.comMutexaGPT: an intuition-to-design translator for physics-based enzyme engineering - Nature Computational ScienceMutexaGPT turns plain-language enzyme-engineering intuitions into lead mutation designs through AI agent-orchestrated high-throughput molecular modeling, yielding experimentally validated improvements...

10 Sept

📢Out now! @mutexazjy.bsky.social and colleagues present MutexaGPT, a multi-agent platform for enzyme engineering that translates intuition into physics-based simulations and thus variant designs. www.nature.com/articles/s43... 🔓 #chemsky

nature.comPredicting brain morphogenesis via physics-transfer learning - Nature Computational ScienceThe authors develop a physics-transfer learning framework that learns cortical folding physics from simple geometries and transfers it to complex brain structures, enabling accurate prediction of brai...

9 Sept

📢New work from Zhiping Xu and colleagues reports a physics-transfer learning framework that enables accurate prediction of brain morphogenesis from limited data. www.nature.com/articles/s43... ⚛️ 🔓 rdcu.be/z1pnN7nqNiia

nature.comAI testing is not AI safety - Nature Computational ScienceNature Computational Science - AI testing is not AI safety

28 Aug

📢In a recent Correspondence, George Breckenridge and Feng Li argue that, while testing is an important component of AI safety governance, it should not be treated as a substitute. www.nature.com/articles/s43... #ArtificialIntelligence 🔓 rdcu.be/XFaZdAqXvzWa

nature.comWhen authors are the best judges of their work - Nature Computational ScienceA large-scale experiment at a major machine learning conference shows that when researchers privately rank their own submissions, those rankings forecast future citations more reliably than peer-revie...

28 Aug

📢Bedoor AlShebli discusses recent work by @wjsu.bsky.social, @kyunghyuncho.bsky.social et al. that finds that authors' rankings of their own AI conference papers predict later citations better than peer-review scores. www.nature.com/articles/s43... #ArtificialIntelligence 🔓 rdcu.be/kFj3SufmJCXa

nature.comEnhancing materials discovery with valence-constrained design in generative modeling - Nature Computational ScienceThis work presents an efficient and modular framework called CrysVCD, which integrates chemical rules directly into the generative process for valence-constrained design.

26 Aug

📢Mingda Li and colleagues develop CrysVCD, a valence-constrained approach for efficient crystal generation. @mitchemistry.bsky.social @mitcheme.bsky.social www.nature.com/articles/s43... #chemsky 🔓 rdcu.be/fCeRM

nature.comLongitudinal alignments and syntheses of multimodal clinical data for personalized medicine with the PULSE framework - Nature Computational SciencePULSE is an AI framework that uses patients’ longitudinal electronic health records to accurately generate within- and cross-modal data from routine clinical laboratory tests, enabling cost-effective precision medicine across diverse biomedical settings.

26 Aug

📢Kang Zhang and colleagues introduce PULSE, an AI framework that uses patients' longitudinal EHRs to accurately generate within and cross-modal data from routine clinical laboratory tests. www.nature.com/articles/s43... 🔓 🖥️ 🧬

nature.comA patient’s longitudinal history reconstructs their unmeasured molecular profile - Nature Computational ScienceMost patients leave the clinic with only a few of the clinical measurements needed to tell the full story of their health. PULSE is a new artificial intelligence (AI) framework that fills in the blank...

26 Aug

An accompanying News & Views by @nrosed.bsky.social and colleagues is also available for this paper! www.nature.com/articles/s43... 🖥️ 🧬 🔓 rdcu.be/fCeK9

nature.comSelf-rankings as a predictor of scientific impact beyond peer review - Nature Computational ScienceThe study finds that authors’ rankings of their own AI conference papers predict later citations better than peer-review scores, suggesting that self-rankings could provide a simple complement to peer...

24 Aug

📢 @wjsu.bsky.social, @kyunghyuncho.bsky.social et al. find that authors' rankings of their own AI papers predict later citations better than peer-review scores, suggesting that self-rankings could complement peer review. www.nature.com/articles/s43... 🔓 #SciencePublishing #ArtificialIntelligence

nature.comResponsible and transparent use of AI in scientific publishing - Nature Computational ScienceAs AI becomes increasingly embedded in research and scientific publishing, transparency, accountability, and human oversight remain essential to safeguarding research integrity.

20 Aug

📢In our latest Editorial, we discuss our policies regarding the appropriate and responsible use of AI. www.nature.com/articles/s43... #SciencePublishing #ResearchPublishing #AcademicPublishing

20 Aug

🚨Our August issue is now live, including quantum solvers for an NP-complete problem, a generative model for information metamaterial design, a method for designing functional RNAs, and much more! Check it out! www.nature.com/natcomputsci... 📰Cover: www.nature.com/articles/s43...

nature.comProvenance grounds trust in autonomous science - Nature Computational ScienceAs large language models begin to plan and run experiments on their own, the reflex is to demand that they be interpretable before they are trusted. We argue that what grounds trust in autonomous scie...

20 Aug

📢In the latest Comment for our five-year anniversary series, @gabegomes.bsky.social and colleagues argue that what grounds trust in autonomous science is not a view inside the model — it's provenance. www.nature.com/articles/s43... #chemsky

nature.comRobust generative transition-state models for unseen chemistry - Nature Computational ScienceAddressing generalization limits in machine learning models for chemical reactions, this work introduces a pretraining scheme that enables accurate predictions for unseen chemistry and expands computa...

13 Aug

📢Out now! @kulikgroup.bsky.social and colleagues introduce targeted benchmarks to probe chemical and structural novelty in generated transition state prediction. www.nature.com/articles/s43... #chemsky 🔓 rdcu.be/fzJeu

nature.comRobust out-of-distribution prediction of Buchwald–Hartwig reactions - Nature Computational ScienceA predictive model capable of generalizing to previously unseen Buchwald–Hartwig chemical space is built to enable rapid, in silico optimization, reducing reaction scouting time from weeks to minutes.

10 Aug

📢New work from @pschwllr.bsky.social and colleagues introduces a model capable of generalizing to unseen Buchwald–Hartwig chemical space, reducing reaction scouting from weeks to minutes. www.nature.com/articles/s43... #chemsky

nature.comModel-guided data production unlocks reliable reaction predictions - Nature Computational ScienceA newly generated high-throughput experimentation (HTE) dataset is integrated with open-source data to produce a large and high-quality Buchwald–Hartwig reaction dataset for training machine learning ...

10 Aug

An accompanying Research Briefing is also available for this paper! www.nature.com/articles/s43... #chemsky 🔓 rdcu.be/fy1vW

nature.comSpatialFormer: universal spatial representation learning from subcellular molecular to multicellular landscapes - Nature Computational ScienceSpatialFormer, 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...

📢Ole Winther and colleagues present SpatialFormer, a transformer-based model for spatial transcriptomics to learn single-cell multimodal and multi-scale information in the niche context. www.nature.com/articles/s43... 🖥️ 🧬 🔓 rdcu.be/fw3dm

nature.comStructure-alignment-driven cross-graph modeling for functional RNA design - Nature Computational ScienceThis study presents AlignIF, which uses multiple structure alignments to capture evolutionary rules and conservation patterns across RNA families, enabling diverse, structure-guided designs of functio...

📢A new study presents a method that uses multiple structure alignments to capture evolutionary rules and conservation patterns across RNA families, enabling diverse, structure-guided design of functional aptamers and ribozymes. www.nature.com/articles/s43... #RNASky 🔓 rdcu.be/fw25L

nature.comStructural alignments to design functional RNAs - Nature Computational ScienceA new method designs RNA sequences by learning from alignments of structurally similar molecules, producing aptamers and ribozymes that are experimentally validated.

An accompanying News & Views for this paper by @ginaelnesr.bsky.social is also available! www.nature.com/articles/s43... #RNASky 🔓https://rdcu.be/fw294

nature.comContinual integration of single-cell multimodal data with MIRACLE - Nature Computational ScienceMIRACLE is an online continual learning framework that efficiently integrates single-cell multimodal data across batches, modalities, tissues, and diseases.

📢Zhen He and colleagues present MIRACLE, an online learning framework for continual integration of single-cell multimodal data. www.nature.com/articles/s43... 🖥️ 🧬

nature.comRethinking privacy-preserving GWAS benchmarks for governance - Nature Computational SciencePrivacy, utility, and efficiency benchmarks have advanced privacy-preserving genome-wide association studies (GWAS) but often fall short of supporting organizational decisions about responsible data s...

23 July

📢In our latest Comment for the 5-year anniversary Series, Natnatee Dokmai argues that, while privacy-preserving GWAS has strong technical benchmarks, they should connect formal guarantees to trust, risk, and governance. www.nature.com/articles/s43... 🖥️ 🧬 🔓 rdcu.be/fvszn

@natrevcomputing.nature.com

23 July

🚨Our July issue is now live, including a model that predicts secondary metabolite chemical structures from biosynthetic gene clusters, a Review on organoid intelligence, and much more! Check it out: www.nature.com/natcomputsci... 📰Cover: www.nature.com/articles/s43...

Nature Computational Science reposted

Nature Reviews Computing

21 July

Our first Collection brings together Reviews, Perspectives and Comments from foundational computer science, applied computing, and intersections with society, highlighting the global, dynamic, and multifaceted nature of the field, and with it the wide range of our scope: go.nature.com/4gFklc4

nature.comCross-task, explainable and real-time decoding of human emotion states by integrating gray and white matter intracranial neural activity - Nature Computational ScienceThis study reports personalized models that decode human emotion states from intracranial brain recordings in real time, work across different tasks and uncover key brain networks that could guide fut...

22 July

📢New Article out today! Yueming Wang and colleagues introduce personalized models that decode human emotion states from intracranial brain recordings in real time. www.nature.com/articles/s43... #compneuro #CogSci 🔓 rdcu.be/fvaLp

nature.comAlphaFold2 turns five - Nature Computational ScienceWe highlight how AlphaFold2 has advanced protein design and inspired the development of artificial intelligence-driven infrastructures for scientific discovery.

15 July

📢AlphaFold2 turns 5 today, and we discuss in our latest Editorial how it has advanced protein design and inspired the development of AI-driven infrastructures. www.nature.com/articles/s43... #alphafold

nature.comUniFFBench: evaluating universal machine learning force fields against experimental measurements - Nature Computational ScienceUniversal machine learning force fields excel on computational benchmarks but consistently fail when tested against experimental measurements of real minerals, underscoring the need for evaluation fra...

14 July

📢Out now! N. M. Anoop Krishnan and colleagues introduce an evaluation framework for assessing the generalization of universal machine learning force fields across the chemical space. www.nature.com/articles/s43... #chemsky 🔓 rdcu.be/ftLgX

nature.comGenerative model for information metamaterial design - Nature Computational ScienceA 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 faste...

14 July

📢Tie Jun Cui and colleagues present a generative model for designing information metamaterials, enabling beam steering, near-field focusing and holography, with more than a 1,000-fold acceleration in holographic design. www.nature.com/articles/s43... 🔓 rdcu.be/ftLeD

nature.comUnderstanding language model scaling for protein fitness prediction - Nature Computational ScienceLarger 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 ach...

13 July

📢Out now! @chaohou.bsky.social, @yshen.bsky.social and colleagues show that PLMs predict fitness best when outputs align with evolutionary patterns in homologs, with the peak performance occurring at moderate predicted sequence likelihoods. www.nature.com/articles/s43... 🔓 rdcu.be/ftBgt

nature.comProtein fitness prediction with language models - Nature Computational ScienceProtein language models are powerful computational tools for protein fitness prediction. The recent investigation into the relationship between model outputs and prediction accuracy reveals the scalin...

13 July

An accompanying News & Views by Tong Wang is also available for this paper! www.nature.com/articles/s43... 🔓 rdcu.be/ftBie