Drainage networks that carry rainwater toward rivers play an important role in floods, but their extent is often difficult to represent in flood models. Researchers from Tokyo University of Science simulated the 2023 flood in Japan's Ichinomiya River Basin using seven drainage network configurations to determine how far upstream channels should be modeled. The findings show that flood models can lose accuracy when drainage networks are either oversimplified or represented in excessive detail.

A Johns Hopkins Medicine-led research team has received an award of up to $52.1 million over four years from the Advanced Research Projects Agency for Health (ARPA-H), part of the Department of Health and Human Services, to better understand the rules that decide what shape a protein takes — and how those rules could support the identification of earlier diagnostics and therapeutic targets when proteins fold the wrong way. The team will start this work, which will be publicly shared through open-access resources, by studying Parkinson’s disease and Ewing sarcoma, a rare cancer of bone and soft tissue in children and young adults.

Improving X-ray imaging under photon-starved conditions could enable lower-dose and higher-resolution imaging for medical diagnosis, security inspection, and industrial testing. Researchers from Chongqing University, the 26th Institute of CETC, and Stevens Institute of Technology developed a scalable photonic crystal scintillator using Marangoni effect-driven interfacial self-assembly. The approach increased average light output by 4.64-fold and improved the X-ray imaging signal-to-noise ratio by 13 dB while maintaining spatial resolution better than 25 μm.

A new review published in PhotoniX Synergy examines how artificial intelligence is transforming the design of terahertz (THz) metamaterial devices (meta-devices). The authors survey conventional and AI-assisted approaches for THz meta-absorbers, modulators, filters, sensors, detectors, and multifunctional devices, and show how machine learning, evolutionary algorithms, deep learning, generative models, and large language models can establish data-driven links between structural geometry and electromagnetic response. The review also discusses challenges including computational cost, limited interpretability, data scarcity, and manufacturability, and outlines future directions such as physics-informed learning, multi-scale optimization, and AI-assisted optical computing hardware for 6G communication, imaging, biosensing, spectroscopy, and security inspection.

Artificial intelligence has already demonstrated great promise in many aspects of biomedical research, from accelerating scientific discovery to speeding the translation of biological insights into new therapies. Scripps Research today announced it has received a significant award of in-kind compute power from the Jen-Hsun and Lori Huang Foundation to advance the adoption of AI into biomedicine and help realize its potential. The grant, totaling more than 14 million hours of compute time, is provided through the Huang Foundation's partnership with CoreWeave using NVIDIA GPUs.

Frontiers’ AI in Science Enablers pledge opens today. The pledge includes AI-powered data curation for 1,000 Horizon Europe datasets and 100 FAIR2 Data Publication Grants for early-career researchers, helping turn research data into peer-reviewed, citable outputs that are better prepared for reuse and AI applications.

Professor Yoonsu Park’s team in the Department of Chemistry automated the entire process from mixing raw materials to synthesis and analysis. The platform continuously produces compounds needed for drug development, showing the potential to increase production volume. The technology is expected to reduce the burden of repetitive experiments and help secure a wider range of drug candidates.