Search

Preferred orientation in cryo-EM often limits structural resolution. Here, the authors present cryoPROS, a novel computational method that co-refines raw and deep generative auxiliary particles to correct misalignment, enabling near-atomic resolution from challenging datasets.

Previous ophthalmic foundation models have struggled to generalize effectively to diverse and rare fundus diseases, restricting their clinical applicability. Here, the authors introduce a vision-language foundation model that demonstrates superior performance in diagnosing both common and rare fundus conditions.

Artificial intelligence (AI) is becoming increasingly relevant to assist with oncology imaging, but diagnosing rare tumours remains challenging. Here, the authors develop an AI approach to detect rare thyroid cancer subtypes by integrating clinical knowledge with image generation based on ultrasound imaging data from large patient cohorts.

The authors present a chip-less wearable sensor-processor integrated neuromorphic system that combines multimodal physicochemical sensing with neuromorphic processing for real-time, autonomous physiological monitoring.

Designing useful molecules demands 3D accuracy and physics awareness. Here, authors present MolEdit, a physics-aligned diffusion model that yields stable structures and enables in-silico molecular editing with chemical and geometric constraints.

Three-dimensional (3D) reconstruction from computed tomography could significantly contribute to guiding lung cancer surgery, but requires comprehensive clinical validation. Here, the authors test the effectiveness of an AI-driven 3D reconstruction system for lung cancer surgery in a retrospective, multi-center, multi-reader, multi-case study.

The NeuroXiv platform enables AI-powered open mining of a database of reconstructions of more than 175,000 mouse neurons, registered to the Common Coordinate Framework. The platform supports queries about connectivity, projection patterns and other morphological features.

Machine learning models represent an opportunity for the automatic generation of histopathology reports. Here, the authors develop HistoGPT, a vision language model that can generate reports from multiple gigapixel-sized whole slide images and also predict tumour thickness, subtypes, and margins, among other diseases.

Unfairness in AI healthcare models arise from biased correlations between sensitive attributes and disease data. Here, the authors show that the Attribute Neutral Framework reduces these biases and enhances fairness without compromising diagnostic accuracy.

DILITracer: an AI-driven ternary hepatotoxicity classification model via spatiotemporal analysis of human liver organoids brightfield images.

DeepRETStroke is a deep learning system using retinal photographs to detect silent brain infarction and predict future stroke events.

Multi-species acute systemic toxicity assessment is fundamental for chemical classification, labelling and risk management. Here, the authors propose a machine learning paradigm, Adjoint Correlation Learning, yielding significant improvements for data-scarce endpoints and offering a useful computational tool for regulatory applications.

In Alzheimer’s disease (AD), disruptions in myelin and axonal structures occur, although the underlying mechanisms remain unclear. Here the authors show that, at the myelin–axon interface, axon–glial signaling, paranodal architecture and amyloid-β aggregation are altered in AD, implicating myelin–axon disruption in disease progression.

Don-Yehiya et al. explore creating an open ecosystem for human feedback on large language models, drawing from peer-production, open-source and citizen-science practices, and addressing key challenges to establish sustainable feedback loops between users and specialized models.

Multimodal large language models are shown to develop object concept representations similar to those of humans. These representations closely align with neural activity in brain regions involved in object recognition, revealing similarities between artificial intelligence and human cognition.

A novel covalent inhibitor, ISM3312, targets the main protease of multiple human coronaviruses, including drug-resistant strains, and shows broad antiviral activity. It offers a promising therapeutic strategy against current and future coronavirus threats.

Artificial Intelligence (AI) models for medical diagnosis often face challenges of generalizability and fairness. Here, the authors show that the Quasi-Pareto Improvement approach is widely applicable to improving AI models among less-prevalent subgroups, promoting equitable healthcare outcomes.

Cancer-targeted drug discovery can be achieved by transcriptomics screening on patients. Here this group reports a drug target screening model built upon triple-negative breast cancer (TNBC) cohort and drug database with the selected drug pair exhibiting effective pyroptosis induction and TNBC tumor growth inhibition.

Chiral materials with strong and tunable chiroptical activities are highly desirable. Here, the authors employ an AI-guided robotic platform to fully execute a cyclic process to perform inverse design and fabricate chiral films with target chiroptical performance.

Early detection and accurate diagnosis of focal liver lesions are crucial for effective treatment and prognosis. Here, the authors present a fully automated diagnostic system that leverages multi-phase CT scans and clinical features, for diagnosing liver lesions.

AI applications on mobile devices are constrained by the high resource demands of large models. Here, the authors developed an efficient on-device multimodal embedding system.

The practicality of memristor-based computation-in-memory (CIM) is limited by the specific hardware design and the manual parameters tuning process. Here, the authors develop a full-stack CIM system with both hardware and software design for improved flexibility and efficiency.

Technical advancements have significantly improved early diagnosis of cervical cancer, but accurate diagnosis is still difficult due to various practical factors. Here, the authors develop an artificial intelligence assistive diagnostic solution to improve cervical liquid-based thin-layer cell smear diagnosis according to clinical TBS criteria in a large multicenter study.

Cytology is commonly used for cervical cancer screening or triage. Here, the authors used deep learning to develop a liquid-based cytology screening approach to detect cervical cancer in both community and hospital-based settings.

Chronic care manages long-term, progressive conditions, while acute care handles short-term ones. Here, authors show chronic conditions account for most of the global health burden, with 68% of DALYs and 93% of YLDs attributed to chronic care needs.

The Einstein Probe has discovered a fast X-ray transient, EP240414a, associated with a broad-lined type Ic supernova. Its unusually soft spectrum and weak jet add to the diversity of stellar deaths and suggest a previously unknown Wolf–Rayet star explosion mechanism.

In a large trial involving 180 classes in different contexts in China, interactions with a chatbot provided to parents increased uptake and appointments for human papillomavirus vaccination, with stronger increases in rural areas.

Artificial-intelligence tools are transforming data-driven science — better ethical standards and more robust data curation are needed to fuel the boom and prevent a bust.

Large-scale irrigation changes local rainfall patterns, leading to more afternoon rain events 10 to 50 km downwind of irrigated areas. These findings will help constrain climate and weather forecasting models and support regional water management.

Nature Computational Science asked a group of scientists to discuss strategies for increasing the presence of Black, Indigenous, People of Color (BIPOC) researchers in computational science, as well as the various considerations to be made for improving education and methods design.

An artificial intelligence (AI) system performs as well as or better than radiologists at detecting breast cancer from mammograms, and using a combination of AI and human inputs could help to improve screening efficiency.

Although theoretical advances in the reactivity of organic compounds have progressed rapidly, a complete understanding of the causal relationships between molecular structures and reaction outcomes based solely on first principles remains elusive. Here, the authors use an in-house HTE platform to conduct over 11,000 distinct acid amine couplings, with an accompanying Bayesian neural network model that gives a prediction accuracy of 89.48% for reaction feasibility.

Ruppin and colleagues overview recent research on the use of AI frameworks in precision oncology, describe ten hallmarks of their contributions across cancer detection, therapy optimization and treatment discovery, and discuss key challenges in clinical implementation

Somatically determined preferential allelic expression of select genes that when mutated cause inborn errors of immunity corresponds with disease phenotypes, suggesting that the penetrance and expressivity of monogenic disorders is also dependent on the ‘transcriptotype’.

An RNA codon-expansion strategy enables incorporation of non-canonical amino acids into proteins of interest orthogonally to existing methods by inserting pseudouridine codons into specific mRNA transcripts and using an engineered decoder tRNA.

Understanding the genetic factors that regulate panicle architecture is important for improving rice yields. In this study, Li et al. demonstrate that the gene PTB1is critical for regulating panicle seed setting rate and pollen tube growth.

The clinical significance of inferring cell spatial profiles from histology images from cancer patients remains to be explored. Here, the authors develop a weakly-supervised deep-learning method, HistoCell, for the direct prediction of super-resolution cell spatial profiles from histology images at the single-nucleus-level.

A memristor-based architecture for federated learning can implement compute-in-memory technology for encryption and decryption computations, a physical unclonable function for key generation and a true random number generator for error polynomial generation all within the same memristor array and peripheral circuits.

Human-machine collaboration in noisy environments using conventional microphones remains challenging. Here, the authors develop an anti-noise triboelectric acoustic sensor integrated with a CNN-based deep learning model to facilitate complex human-machine tasks in highly noisy conditions.

Resource scarcity and associated trade risks will limit China’s comprehensive energy transition; even with possible material substitution, the contribution of wind and solar power will be far from expected, yielding a big gap of carbon reductions.