Researchers discovered that embryos must clear maternal mRNAs to develop properly. When cleanup fails, it causes aberrant transcription and R-loop formation, which blocks DNA replication, damages the genome, and arrests development. Resolving R-loops can reverse these defects, revealing a critical quality-control mechanism at life's beginning.

Researchers have developed miniature magnetic robots that mimic fish behavior, working together as coordinated swarms to deliver drugs precisely and efficiently to tissue. The breakthrough could transform treatment of conditions where individual tiny robots lack sufficient coverage area for effective therapy.

With the rapid advancements in computer technology and bioinformatics, the prediction of protein-ligand binding sites has become a central component of modern drug discovery and development. Traditional experimental methods are often constrained by long experimental cycles and high costs; therefore, the development of accurate and efficient computational methods is of paramount significance for conserving time and cost. This review comprehensively summarizes the methodological advancements and current applications in the field of screening for druggable protein target sites, systematically comparing the fundamental principles, advantages, and disadvantages of four main categories of methods: structure- and sequence-based methods, machine learning-based methods, binding site feature analysis methods, and druggability assessment methods. Subsequently, by integrating classic case studies, this paper elaborately discusses the technical support and theoretical guidance afforded by the screening of protein druggable target sites for drug discovery and drug repositioning. Finally, this paper thoroughly explores the current challenges inherent in the field of protein-ligand binding site prediction, with a particular focus on future technological trends, systematically elucidating the developmental prospects and potential applications of these predictive methods.

Large language models (LLMs) are increasingly integrated into oncology care, but their ability to maintain accurate performance over time remains poorly understood. A new study evaluates the temporal evolution of three leading LLMs—ChatGPT-3.5, ChatGPT-4, and Gemini—using 614 oncology-related questions (223 subjective, 391 multiple-choice) from 23 studies. Results show ChatGPT-3.5 and ChatGPT-4 exhibit declining accuracy over time, while Gemini demonstrates significant improvement. Notably, ChatGPT-3.5’s performance on subjective questions shifted from superior to inferior relative to original queries between March and April 2023, with the gap widening. These findings highlight the critical need for ongoing monitoring of LLMs in clinical oncology settings to ensure reliable, up-to-date support for healthcare professionals and patients.

The Chinese Alliance of Research for Mesothelioma (ChARM) has released China’s first expert consensus on Malignant Mesothelioma of the Tunica Vaginalis Testis (MMTVT). A rare malignancy accounting for 0.3%–5% of all malignant mesotheliomas, MMTVT has long suffered from misdiagnosis and inconsistent treatment due to nonspecific symptoms and lack of unified guidelines. The consensus integrates global evidence and national expert experience, covering epidemiology, diagnosis, treatment, prognosis, and follow-up. It emphasizes multidisciplinary collaboration (MDT), stratified screening, and personalized therapy, providing authoritative clinical guidance to enhance patient care.

A recent study published in National Science Review, reveals that continuously increasing greenhouse gases (GHG) emissions will significantly amplify the risk of extreme dry-hot in North America and Europe by enhancing land-air coupling. This study highlights the critical role of regional climate feedbacks under global warming.