Search

This study identifies key neurocognitive domains that distinguish patients with schizophrenia from healthy individuals using machine learning. Analyzing data from 1,304 participants, it demonstrates that verbal learning and emotion identification effectively classify conditions, promoting efficient neurocognitive profiling strategies.

Genome-wide association analyses using data from 19 biobanks identify variants influencing risk of thyroid diseases and yield polygenic risk scores associated with features of aggressive thyroid cancer.

This study presents MAOSS, a multimodal AI model that repurposes non-contrast CT scans and leverages clinical features to detect and stage liver steatosis and fibrosis. Here the authors show MAOSS accurately stratifies cirrhosis progression risk when embedded into the standard clinical workflow, enabling scalable, opportunistic screening for early intervention of steatotic liver disease.

Primary angle-closure glaucoma can cause blindness. Here, the authors study populations of Asian and European ancestries and find that genetic factors associated with darker irises and being far sighted increase the risk of developing the condition.

Artificial intelligence applied to brain magnetic resonance imaging could transform neurological care. Here the authors present BrainIAC, a foundation model that can be used across diverse clinical tasks with limited training data.

This study identifies protein signatures of pediatric obesity linked to cardiometabolic risk, shows treatment-related changes, and highlights a three-protein panel predictive of steatotic liver disease for early diagnosis

The authors show that phosphorylated tau 217 is an accurate blood-based biomarker to detect Alzheimer’s disease and predict disease progression in individuals with Down syndrome, supporting its use as a reliable, non-invasive tool for early diagnosis and care.

A streamlined blood test using mass spectrometry improves measurement of amyloid-β for early Alzheimer’s disease diagnosis, requiring less sample volume and reagents while maintaining high accuracy, sensitivity and strong agreement with brain imaging.

The authors show that plasma AT(N) biomarkers can distinguish Alzheimer’s disease and frontotemporal lobar degeneration in diverse Latin American populations. Using machine learning and integrating neuroimaging, significant diagnostic accuracy was achieved, enhancing clinical assessments of these conditions in Latin America.

A comprehensive benchmarking of 19 histopathology foundation models finds that a vision-language foundation model outperforms vision-only models.

EGFR mutations are common in non-small cell lung cancer and patients with these mutations are treated with tyrosine kinase inhibitors. Here, the authors show that EGFR mutation status can be predicted from ¹⁸F-FDG-PET/CT images, which may enable the stratification of patients for treatment.

Lan et al. demonstrate the high performance of plasma p-tau217 in detecting and tracking multiple aspects of Alzheimer’s disease pathology, outperforming other plasma tau biomarkers across two Chinese cohorts (GHABS and Huashan) and the North American ADNI cohort.

Deep learning methods can be a powerful part of digital pathology workflows, provided well-annotated training datasets are available. Tolkach and colleagues develop a deep learning model to recognize and grade prostate cancer, based on a convolution neural network and a dataset with high-quality labels at gland-level precision.

Multi-trait genome-wide analyses identify variants associated with comorbid lung diseases. Polygenic scores leveraging shared components of heritable risk improve prediction of asthma, chronic obstructive pulmonary disease and lung cancer in a multi-ancestry cohort.

Delphi-2M forecasts a person’s future health, covering more than 1,000 diseases, provides insights into co-morbidity dynamics and generates synthetic data for the training of AI models that have never seen actual data.

Trained on 1.5 million whole-slide images from 100,000 patients, a pathology foundation model is shown to improve performance of specialized models in detection of rare cancers.

Traditional type 2 diabetes (T2D) screening methods often depend on age, BMI guidelines and glucose measurements. Here, authors use a deep learning model that leverages chest radiographs and electronic health record data to screen for T2D, highlighting potential for early detection and intervention.

Occult node metastasis is a key staging component of non-small cell lung cancer. Here, the authors use deep learning to improve diagnosis of lymph node metastasis from PET and CT radiomics.

Among 1,767 patients in 5 centers, a fully automated blood test showed high accuracy for Alzheimer’s pathology in secondary and primary care using a predefined cutoff.

This multicenter study demonstrates use of dried and capillary blood as a minimally invasive, scalable approach for Alzheimer’s biomarker testing in research, with potential as a widely scalable population-based research approach, especially in resource-limited settings.

Machine learning is increasingly used to support IVF pretreatment prognostic counseling. Here, the authors find that center-specific models improved IVF live birth predictions over a US registry model, and draw conclusions for improving IVF access, affordability and patient-centric counseling.

Cancer biomarkers are often continuous measurements, which poses challenges for their prediction using classification-based deep learning. Here, the authors develop a regression-based deep learning method to predict continuous biomarkers - such as the homologous repair deficiency score - from cancer histopathology images.

Atrial fibrillation has a strong genetic basis, but key mechanisms remain unclear. Here, the authors show that a cross-population GWAS meta-analysis identifies 525 loci and highlights shared and ancestry-specific pathways relevant to AF risk and therapeutic targeting.

Chang et al. performed a pan-cancer multimodal data integration analysis and devised a model, LORIS, that can predict objective responses to immunotherapy and patient survival across many cancer types and allow for patient stratification.

Researchers identified three distinct white matter hyperintensity progression patterns, validated in both patients and healthy individuals. Each pattern links to different vascular risks and outcomes, improving stroke prediction.

Polygenic scores for body mass index and obesity constructed using data from 5 million people with different ancestries were associated with distinct weight gain trajectories from early childhood to adulthood.

The function and regulatory mechanism of mRNA acetylation modification in cancer progression remains largely unknown. Here the authors identify that targeting tumor N4-acetylcytidine (ac4C) RNA acetyltransferase NAT10 improves anti-tumor response via MYC/CDK2/DNMT1/dsRNA/type I IFN pathway, and show the synergic anti-tumor effect of NAT10 inhibition with aPD-1.

A range of blood-based biomarkers have shown high specificity for Alzheimer’s disease (AD) pathophysiology with phosphorylated-tau (p-tau) being the most promising test. Here, the authors show the utility of plasma p-tau212 in autopsy-confirmed AD and memory clinic patient cohorts.

Immune checkpoint inhibitor treatment improves overall survival in metastatic urothelial cancer, but response of individual patients varies significantly. Authors here perform whole-genome DNA and bulk RNA sequencing on samples from metastatic tumours and based on these data, they set up a single metric, T cell-to-stroma enrichment (TSE) score, that reflects the relative abundance of T cells versus stromal cells and their products, accurately predicting therapeutic outcome.

Here, Li et al. integrate population genetic and protein structural perspectives to map the landscape of 3D constraint on 80% of human proteins. They show that 3D mutational constraint is pervasive and strongly associated with functional relevance.

Here, by integrating faecal metabolomics, metagenomics, and habitual dietary data of two large human cohorts, the authors show that faecal metabolites reflect diet and gut microbiome interactions, predict dietary patterns, and indicate cardiovascular risk, offering insights for diet-based health interventions.

Large-scale genome-wide analyses identify hundreds of genetic loci associated with hypothyroidism and thyroid hormone levels, demonstrating the potential of using polygenic risk scores to predict disease onset and progression.

Going from model development to a pilot implementation study, a deep learning model shows that acute aortic syndrome can be diagnosed directly from noncontrast CT, increasing accuracy and decreasing time to diagnosis.

Plasma P-tau, in combination with clinical measures, predicts future Alzheimer’s disease dementia in two independent cohorts with high accuracy and is superior to the clinical diagnostic predictions of specialists.

Prognostic markers for patients with COVID-19 are of critical importance in determining the course of SARS-CoV-2 infection and patient handling. Here the authors determine and apply a prognostic proteomic panel for risk and drug prediction in the management of SARS-CoV-2 infected patients.