Open Materials 2024 dataset enables machine learning models to surpass previous standards, achieving F1 scores above 0.9 and 20 meV per atom accuracy in inorganic materials.
- 18 cites
Nature Computational Science
Nature Portfolio · Genomics & Bioinformatics · ISSN 2662-8457
- Impact (2-yr)
- 11.2
- h-index
- 77
- Open access
- 37%
- Works
- 1.1K
- Papers this month
- 8
Rebuilt
Moved this week
NEP89 delivers high-accuracy, computationally efficient atomistic simulations across 89 elements, enabling large-scale studies previously limited by resource demands.
- Open access
- 28 cites
Latest in Nat Comput Sci
Newest first · last 60 days
Modulating complex oscillators to achieve multiple properties is feasible with a new framework that ensures parameter uniqueness and efficiency, validated across diverse systems.
Large language models can serve as human proxies in various scientific and applied contexts, each role requiring distinct validity criteria.
MutexaGPT enables physics-based enzyme design from plain-English intuition, generating actionable mutation proposals with high-throughput molecular modeling in diverse engineering tasks.
Physics-transfer learning enables accurate prediction of brain morphogenesis with limited labeled data, achieving strong generalization across complex biological systems.
- Open access
Algorithmic curation's impact on online content consumption is inconclusive, with human preferences playing a significant role alongside algorithms.
A new framework improves out-of-distribution prediction accuracy for Buchwald-Hartwig reactions, achieving reliable results on unseen substrates and conditions.
MIRACLE enables scalable, accurate continual integration of single-cell multimodal data, expanding biological atlases with new datasets across tissues and diseases.
Local pseudopotentials enhance the efficiency and accuracy of neural network-based quantum Monte Carlo for large, complex systems, including the Fe4S4 cluster.
PULSE accurately reconstructs metabolomic and proteomic profiles from sparse blood tests, outperforming benchmarks and enabling disease prediction with AUROCs up to 0.83.
SpatialFormer achieves accurate spatial gene expression modeling by integrating multimodal data, enabling comprehensive analysis across single-cell and multicellular systems.
TDiMS descriptor outperforms existing methods by accurately capturing nonlocal intramolecular interactions with 20% higher predictive accuracy in property prediction tasks.
- Open access
PathSegmentor, a foundation model using natural language, achieves top performance in pathology image segmentation across diverse datasets with 275,200 image-mask-label triples.
Authors' self-rankings of their AI conference submissions predict future citations twice as well as peer review scores, highlighting their value in assessing research impact.
Valence-constrained generative model CrysVCD achieves 85% metastability in crystal design, significantly improving chemical validity in materials discovery.
Self-supervised pretraining enhances generative transition-state models, reducing median RMSD by over 50% on unseen transition metal complexes in Transition1x-TMC reactions.
Follow Nat Comput Sci in the app
Browse JClub
Journals this month
- bioRxiv5569
- medRxiv1746
- Environ Res485
- PNAS463
- Sci Adv439
- JAMA255
- Cell Rep210
- BMJ205
- Nat Commun177
- Environ Pollution170
- Lancet149
- Science138
- NEJM119
- Applied & Environ Microbio116
- Curr Biol116
- J Exp Bot115
- mBio109
- Nature104
Moving areas, week to 3 Oct 2026
- Single-cell and spatial transcriptomics10
- Artificial Intelligence in Healthcare and Education8
- CAR-T cell therapy research7
- Neuroinflammation and Neurodegeneration Mechanisms6
- Lung Cancer Treatments and Mutations5
- Diabetes Treatment and Management4
- Pancreatic and Hepatic Oncology Research4
- Photosynthetic Processes and Mechanisms4
- Genomics and Chromatin Dynamics4
- Gut microbiota and health4