Search

Humanity’s Last Exam, a multi-modal benchmark at the frontier of human knowledge, is designed to be an expert-level closed-ended academic benchmark with broad subject coverage.

By combining satellite observations with ground-based data and expert validation, this analysis demonstrates considerable misestimation of grassland extent and thereby carbon stock estimates in previous global assessments based on remote sensing.

A prospective nationwide genomic screening pilot for ten genes in young adults in Australia detected pathogenic variants in 2% of the population, most of whom were ineligible for government-funded genomic testing.

Global Burden of Disease estimates show that between 1990 and 2023, the prevalence and burden of drug use disorders, inclusive of amphetamine, cannabis, cocaine and opioid use, have been increasing in high-income countries, particularly in the USA.

Nudt21-mediated 3′UTR regulation orchestrates tissue regeneration through dose-dependent control of stem cell differentiation and multiprotein complex assembly, revealing alternative polyadenylation as essential for organismal survival.

An expert recommendation is presented for the design and evaluation of research studies using transcranial electrical stimulation, resulting in the selection of a 66-point checklist aimed at improving the quality of transcranial electrical stimulation studies.

Vassy, Dornisch and colleagues developed a genomics-based prostate cancer risk model to support a randomized clinical trial of precision screening in a national healthcare system.

An annotation-free vision–language model based on contrastive learning performs pathology localization and clinical diagnosis on a wide variety of medical image datasets.

Washington

Opponents of moves to limit global warming under the Kyoto Protocol have seized upon a controversial new scientific paper that appears to absolve carbon dioxide of blame for the rapid warming seen over the past few decades.

In a multicenter, randomized trial of patients with atrial fibrillation and a low risk of thromboembolic events, treatment with the anticoagulant rivaroxaban showed no benefit in reducing cognitive decline, stroke or transient ischemic attack when compared to placebo.

Rathmell and colleagues show that metabolic reprograming of regulatory T cells is associated with severity of critical illness in patients with and without sepsis.

Estimates from the Global Burden of Disease study reveal declines in chronic respiratory disease mortality between 1990 and 2023—especially during the COVID-19 pandemic—but the burdens on people affected remain substantial.

Bruijns et al. present a modeling tool that enables the tracking of learning dynamics across subjects to reveal how behaviors emerge and adapt. Applying the tool to a decision-making task in mice uncovers similarities and differences across individuals.

The physiological phenomenon whereby heart rate varies in phase with breathing in vertebrates has been known as ‘respiratory sinus arrhythmia’. In this Expert Recommendation, the authors argue that this terminology is potentially misleading and propose replacing it with the term ‘respiratory heart rate variability’, which avoids pathological connotations.

Citizens’ approval is crucial for maintaining aid to Ukraine’s defense against Russian aggression. Here, the authors show that Western citizens back Ukraine’s fight, but moral and strategic concerns, as well as internal divisions, may impose restraints.

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.

Pl@ntBERT is a language-based AI model that learned the ‘syntax’ of plant assemblages, predicting likely species and inferring habitats by modelling biotic relationships.

Lung adenocarcinomas bearing the ID2 mutational signature display increased LINE-1 retrotransposon activity, which contributes to their fast evolutionary dynamics and aggressive phenotype.

Large language models can describe and classify changing objects in astronomical images with high accuracy. This enables searches for visual features using text and introduces a new way to interact with sky survey data.

EchoPrime, a video-based foundation artificial intelligence model for echocardiography, is presented.

Freezing of gait (FOG) is among the most debilitating symptoms of Parkinson disease. This Consensus Statement from the International Consortium for Freezing of Gait presents new guidelines for the definition and assessment of FOG, with the aim of harmonizing the study and management of the condition.

FLOWERING LOCUS M (FLM) is known as a repressor of Arabidopsis flowering. Here, the authors show that a single intronic substitution of FLM modulates leaf color and plant growth strategy along the leaf economics spectrum, as well as plays a role in plant adaptation.

The Taiwan Precision Medicine Initiative (TPMI) has built a cohort of more than half a million individuals with Han Chinese ancestry, providing a rich resource of genetic and medical data for this group.

BRCA1 and BRCA2 are well known breast cancer predisposition genes, however, many variants have not yet been classified for their pathogenicity. Here, the authors analyse a large combined cohort to provide evidence of variant pathogenicity.

Concerns about energy and transport poverty are critical to the net-zero transition but are growing in the wake of recent global events. This study presents findings from focus groups with the public and expert interviews in the United Kingdom on support for different policy options to address these challenges.

Clonalscope is a method for cancer subclone detection leveraging copy number profiles estimated using spatial and single-cell sequencing data.

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.

Humans often interact without knowing the cooperative or competitive intentions of others. Here, the authors determined the neurocomputational mechanisms engaged in adapting to fluctuating intentions of others over repeated social interactions.

An important problem in microscopy analysis is to reduce the need for manual annotation of datasets. Bilodeau et al. develop a neural network method that can be trained on a simple main classification task using image-level annotations to solve more complex tasks such as semantic segmentation.

Cue reactivity measured by functional magnetic resonance imaging is used in studies of substance-use disorders. This Consensus Statement is the result of a Delphi process to arrive at parameters that should be reported in describing these studies.

RFdiffusion2, an extension of the RFdiffusion framework, builds de novo enzyme active sites using atom-level functional group constraints.

The study advances the use of serological surveys to guide trachoma elimination program decisions and provides a way to set thresholds for whether or not to continue an intervention program.

Guidance is lacking on how to best integrate sex, gender and social and structural determinants of health into neuroscience research on brain resilience in ageing and dementia. In this Roadmap article, Rajah et al. propose a way forward for conducting more inclusive research in this field.

Optimizations of X-ray nanotomography including the choice of resin allows high-resolution imaging of mouse brain tissue, approaching the resolution of volume electron microscopy. Since it does not require slicing the tissue, this technology may become an attractive alternative to current standard methods for connectomics.

Together with a companion paper, the generation of a transcriptomic atlas for the mouse lemur and analyses of example cell types establish this animal as a molecularly tractable primate model organism.

The second Global Amphibian Assessment finds that the status of amphibians is continuing to deteriorate globally, driven predominantly by climate change, disease and habitat loss.

Pretrained using 335,645 whole-slide images, a foundation model is developed to provide representations for slide- and patient-level tasks. It is capable of performing clinical tasks and generating reports even in data-scarce scenarios, such as rare cancer diagnosis and survival prediction, without requiring further fine-tuning.

Stressors such as sediment dredging can harm marine organisms, but this impact could be minimised if targeted within ‘ecological windows’. Here, Wu and colleagues develop a modelling framework to identify ecological windows that maximise seagrass resilience under varying dredging schedules.

Citizen science taps the efforts of non-experts. Here, authors describe Drugit, an extension of the crowdsourcing game Foldit, and its use in designing a non-peptide binder of Von Hippel Lindau E3 ligase for use with proteolysis targeting chimeras.

An implantable vagus nerve-targeted device safely reduces disease activity and joint damage in rheumatoid arthritis, offering a new nondrug treatment for patients who do not respond to or cannot tolerate medication.