OpenML: Insights from 10 years and more than a thousand papers.

OpenML is an open-source platform that democratizes machine-learning evaluation by enabling anyone to share datasets in uniform standards, define precise machine-learning tasks, and automatically share detailed workflows and model evaluations. More than just a platform, OpenML fosters a collaborative ecosystem where scientists create new tools, launch initiatives, and establish standards to advance machine learning. Over the past decade, OpenML has inspired over 1,500 publications across diverse fields, from scientists releasing new datasets and benchmarking new models to educators teaching reproducible science.

From histology to diagnosis: Leveraging pathology foundation models for glioma classification.

The fifth edition of the WHO classification of brain tumors increasingly emphasizes the role of extensive genetic testing in the diagnosis of gliomas. In this context, computational pathology foundation models (FMs) present a promising approach for inferring molecular entities directly from conventional, H&E-stained histological images, potentially reducing the need for genetic analysis. We conducted a robust investigation into the ability of five established FMs to generate effective embeddings for downstream glioma classification using three datasets (TCGA, n=839 samples; EBRAINS, n=786 samples; TUM, n=250 samples) and state-of-the-art augmentation techniques.

Flashzoi: An enhanced Borzoi for accelerated genomic analysis.

Motivation: Accurately predicting how DNA sequence drives gene regulation and how genetic variants alter gene expression is a central challenge in genomics. Borzoi, which models over ten thousand genomic assays including RNA-seq coverage from over half a megabase of sequence context alone promises to become an important foundation model in regulatory genomics, both for massively annotating variants and for further model development. However, the currently used relative positional encodings limit Borzoi's computational efficiency.

Comparing supervised machine learning algorithms for the prediction of partial arterial pressure of oxygen during craniotomy.

Background And Objectives: Brain tissue oxygenation is usually inferred from arterial partial pressure of oxygen (paO), which is in turn often inferred from pulse oximetry measurements or other non-invasive proxies. Our aim was to evaluate the feasibility of continuous paO prediction in an intraoperative setting among neurosurgical patients undergoing craniotomies with modern machine learning methods.

Methods: Data from routine clinical care of lung-healthy neurosurgical patients were extracted from databases of the respective clinical systems and normalized.

TDEase: An Open-Source Data Visualization Software Framework for Targeted Proteoform Characterization by Top-Down Proteomics.

Top-down proteomics (TDP) is a powerful approach for characterizing intact protein molecules and their diverse proteoforms. Despite recent advances, current TDP software tools often suffer from fragmented workflows, steep learning curves for non-experts, or limited interactive visualization capabilities. To address these challenges, we introduce TDEase, an integrated analytical framework designed to streamline and enhance TDP data interpretation, with a current focus on integration with the TopPIC suite package for targeted proteoform characterization.

Addressing data gaps in sustainability reporting: A benchmark dataset for greenhouse gas emission extraction.

Reliable company-level greenhouse gas (GHG) emissions data are essential for stakeholders addressing the climate crisis. However, existing datasets are often fragmented, inconsistent, and lack transparent methodologies, making it difficult to obtain reliable emissions data. To address this challenge, we present a gold standard dataset containing emission metrics extracted from 139 sustainability reports collected from company websites.

Ab Initio Accuracy Neural Network Potential for Drug-Like Molecules.

The advent of machine learning (ML) in computational chemistry heralds a transformative approach to one of the quintessential challenges in computer-aided drug design (CADD): the accurate and cost-effective calculation of atomic interactions. By leveraging a neural network (NN) potential, we address this balance and push the boundaries of the NN potential's representational capacity. Our work details the development of a robust general-purpose NN potential, architected on the framework of DPA-2, a deep learning potential with attention, which demonstrates remarkable fidelity in replicating the interatomic potential energy surface for drug-like molecules comprising 8 critical chemical elements: H, C, N, O, F, S, Cl, and P.

Specialized curricula for training vision language models in retinal image analysis.

Clinicians spend significant time reviewing medical images and transcribing findings. By integrating visual and textual data, foundation models have the potential to reduce workloads and boost efficiency, yet their practical clinical value remains uncertain. In this study, we find that OpenAI's ChatGPT-4o and two medical vision-language models (VLMs) significantly underperform ophthalmologists in key tasks for age-related macular degeneration (AMD).

Speckle2Self: Self-supervised ultrasound speckle reduction without clean data.

Image denoising is a fundamental task in computer vision, particularly in medical ultrasound (US) imaging, where speckle noise significantly degrades image quality. Although recent advancements in deep neural networks have led to substantial improvements in denoising for natural images, these methods cannot be directly applied to US speckle noise, as it is not purely random. Instead, US speckle arises from complex wave interference within the body microstructure, making it tissue-dependent.

Traffic-Emitted Amines Promote New Particle Formation at Roadsides.

New particle formation (NPF) is a major source of atmospheric aerosol particles, significantly influencing particle number concentrations in urban environments. High condensation and coagulation sinks at highly trafficked roadside sites should suppress NPF due to the low survival probability of clusters and new particles, however, observations show that roadside NPF is frequent and intense. Here, we investigate NPF at an urban background and roadside site in Central Europe using simultaneous measurements of sulfuric acid, amines, highly oxygenated organic molecules (HOMs), and particle number size distributions.

Orthrus: a Pumilio-family gene involved in fruiting body and dark stipe development in .

Fruiting bodies of mushroom-forming fungi (Agaricomycetes) are complex multicellular structures whose formation is regulated by a developmental program that dynamically responds to environmental changes, such as light intensity. However, the genetic architecture and regulation of this developmental program are poorly known. Here, we characterize a novel Pumilio family gene, , which influences fruiting body development, particularly the formation of dark stipes, a light-dependent alternative developmental trajectory.

V2C-Long: Longitudinal cortex reconstruction with spatiotemporal correspondence.

Reconstructing the cortex from longitudinal magnetic resonance imaging (MRI) is indispensable for analyzing morphological alterations in the human brain. Despite the recent advancement of cortical surface reconstruction with deep learning, challenges arising from longitudinal data are still persistent. Especially the lack of strong spatiotemporal point correspondence between highly convoluted brain surfaces hinders downstream analyses, as local morphology is not directly comparable if the anatomical location is not matched precisely.

Single-cell differential expression analysis between conditions within nested settings.

Differential expression analysis provides insights into fundamental biological processes and with the advent of single-cell transcriptomics, gene expression can now be studied at the level of individual cells. Many analyses treat cells as samples and assume statistical independence. As cells are pseudoreplicates, this assumption does not hold, leading to reduced robustness, reproducibility, and an inflated type 1 error rate.

Deformable image registration of dark-field chest radiographs for functional lung assessment.

Background: Dark-field radiography of the human chest has been demonstrated to have promising potential for the analysis of the lung microstructure and the diagnosis of respiratory diseases. However, most previous studies of dark-field chest radiographs evaluated the lung signal only in the inspiratory breathing state.

Purpose: Our work aims to add a new perspective to these previous assessments by locally comparing dark-field lung information between different respiratory states to explore new ways of functional lung imaging based on dark-field chest radiography.

Attitudes, Perceptions, and Factors Influencing the Adoption of AI in Health Care Among Medical Staff: Nationwide Cross-Sectional Survey Study.

Background: Artificial intelligence (AI) has demonstrated transformative potential in the health care field; yet, its clinical adoption faces challenges such as inaccuracy, bias, and data privacy concerns. As the primary operators of AI systems, physicians and nurses play a pivotal role in integrating AI into clinical workflows. Their acceptance and use of AI are essential for bridging the gap between technological innovation and practical implementation.

Deep learning for sub-ångström-resolution imaging in uncorrected scanning transmission electron microscopy.

Achieving sub-ångström resolution has long been restricted to sophisticated aberration-corrected scanning transmission electron microscopy (AC-STEM). Recent advances in computational super-resolution techniques, such as deconvolution and electron ptychography, have enabled uncorrected STEM to achieve sub-ångström resolution without the need for delicate aberration correctors. However, these methods have strict requirements for sample thickness and thus have yet to be widely implemented.

Altered translation elongation contributes to key hallmarks of aging in the killifish brain.

Aging is a major risk factor for neurodegeneration and is characterized by diverse cellular and molecular hallmarks. To understand the origin of these hallmarks, we studied the effects of aging on the transcriptome, translatome, and proteome in the brain of short-lived killifish. We identified a cascade of events in which aberrant translation pausing led to altered abundance of proteins independently of transcriptional regulation.

Towards quantitative intensity analysis of conventional T1-weighted images in multiple sclerosis.

Conventional T1-weighted (T1w) magnetic resonance imaging (MRI) is commonly used in multiple sclerosis (MS) morphometry and volumetry research. However, arbitrary intensity scales preclude interpretation of signal values across patients, sites, and time. This requires quantitative MRI techniques, which are not always available.

Predictive modeling for step II therapy response in periodontitis - model development and validation.

Steps I and II periodontal therapy is the first-line treatment for periodontal disease, but has varying success. This study aimed to develop machine learning models to predict changes in periodontal probing depth (PPD) after step II therapy using patient-, tooth-, and site-specific clinical covariates. Models accurately predicted that healthy sites stay healthy, but performed suboptimally for diseased sites.

Outlier Detection in Mendelian Randomization.

Mendelian randomization (MR) uses genetic variants as instrumental variables to infer causal effects of exposures on an outcome. One key assumption of MR is that the genetic variants used as instrumental variables are independent of the outcome conditional on the risk factor and unobserved confounders. Violations of this assumption, that is, the effect of the instrumental variables on the outcome through a path other than the risk factor included in the model (which can be caused by pleiotropy), are common phenomena in human genetics.

Predicting intersystem crossing rate constants of alkoxy-radical pairs with structure-based descriptors and machine learning.

Peroxy radicals (RO) are ubiquitous intermediates in many oxidation processes, especially in the atmospheric gas phase. The recombination reaction of two peroxy radicals (RO + R'O) has been demonstrated to lead, several steps, to a triplet complex of two alkoxy radicals: (RO˙⋯R'O˙). The different product channels of RO + R'O reactions thus correspond to different reactions of this triplet complex.

Facial Emotion Recognition of 16 Distinct Emotions From Smartphone Videos: Comparative Study of Machine Learning and Human Performance.

Background: The development of automatic emotion recognition models from smartphone videos is a crucial step toward the dissemination of psychotherapeutic app interventions that encourage emotional expressions. Existing models focus mainly on the 6 basic emotions while neglecting other therapeutically relevant emotions. To support this research, we introduce the novel Stress Reduction Training Through the Recognition of Emotions Wizard-of-Oz (STREs WoZ) dataset, which contains facial videos of 16 distinct, therapeutically relevant emotions.

Beyond the label "major depressive disorder"-detailed characterization of study population matters for EEG-biomarker research.

Introduction: Major Depressive Disorder (MDD) is a prevalent, multi-faceted psychiatric disorder influenced by a plethora of physiological and environmental factors. Neuroimaging biomarkers such as diagnosis support systems based on electroencephalography (EEG) recordings have the potential to substantially improve its diagnostic procedure. Research on these biomarkers, however, provides inconsistent findings regarding the robustness of specific markers.

Navigating disruption in the PID landscape: embracing opportunities and anticipating threats in the next ten years.

Introduction: The International Patient Organisation for Primary Immunodeficiencies (IPOPI) held its third edition of the Global Multi-Stakeholders' Summit, gathering key primary immunodeficiencies (PID) stakeholders and experts to discuss and foment global collaboration.

Methods: This edition focused on the impact of genomic medicine in PID treatment, the role of digital health, including artificial intelligence, in PID care, and how to anticipate and minimise risks to ensure optimal patient access to care.

Results: These discussions aimed to examine current hurdles and brainstorm feasible solutions and priorities for the PID community in these areas in the next ten years.

Positional BMP signaling orchestrates villus length in the small intestine.

The intestinal epithelium undergoes fast turnover, and the villus length in the small intestine gradually decreases from the duodenum to the ileum. However, the underlying mechanisms remain poorly understood. In this study, we investigate the regulatory mechanism underlying the regional disparity of villus length.