Mendelian randomization leverages genetic variants to mitigate confounding bias in causal inference. This paper systematically reviews the methodological evolution of Mendelian randomization, elaborates on its expanding applications in epidemiology and drug target validation, and identifies future research directions to address existing biases via dynamic analysis, multi-omics integration, and trans-ethnic data harmonization.

This review summarizes the immuno-inflammatory mechanisms driving the pathogenesis of osteoarthritis (OA), extending beyond the traditional view of OA as a purely degenerative disease to incorporate immune cell infiltration, inflammatory factor release, and the formation of a chronic low-grade inflammatory microenvironment. Focusing on the activation and dysregulation mechanisms of innate and adaptive immunity in OA, we delineate recent advances in understanding OA immunopathology and discuss potential therapeutic strategies targeting immune responses, aiming to provide new insights for the prevention and treatment of OA.

Wireless charging has already transformed how people power phones, wearables and even electric vehicles. The next leap is advancing wireless technologies to not merely deliver electrical energy, but to enable direct non-electrical outputs, such as optical, thermal and mechanical energy, directly according to actual needs, building an energy delivery network as flexible and efficient as the internet. To drive this transformation, researchers at The Hong Kong Polytechnic University (PolyU) have conducted a groundbreaking study that sets out a technical roadmap for “wireless energy internet”, guiding the industry from “wireless power transfer” toward direct “wireless energy conversion”.

In today's digital world, everything from medical images and financial records to personal photos and secure communications depends on encryption systems that can protect sensitive information. But real-world technologies rarely operate under perfect conditions. Measurements can be imprecise, hardware introduces small errors, and communication channels are often affected by noise.

Four decades after the 1986 nuclear disaster, researchers at Leibniz University Hannover and the Helmholtz-Zentrum Dresden-Rossendorf (HZDR) examined six tiny, highly radioactive fragments that were released from the destroyed reactor in Ukraine in the wake of the incident. Analyses show that these “hot particles” are far more stable than had previously been assumed. The findings could allow for more precise assessments of the health risks posed by such radioactive particles (DOI: 10.1016/j.jhazmat.2026.141533).

A research team has found that warming changes where nitrogen is held in flooded rice soil.

Researchers at Universidad Carlos III de Madrid (UC3M) have designed and tested a new type of modular tire aimed at making it smarter and more sustainable. As part of the ECOTIRE project, they have developed a removable silicone tread that can significantly increase durability, reduce abrasive wear, and mitigate pollution caused by rubber waste.

A research team has revealed that different surface oxygen species determine how effectively platinum-based catalysts release hydrogen from aqueous methanol.

Deep learning overcomes the limitations of traditional methods in behavioral prediction and neural decoding—such as handcrafted features, linear models, and physical priors—and through high-dimensional nonlinear modeling, multimodal fusion, and interpretable learning, significantly improves the accuracy of action and trajectory prediction as well as the mechanistic interpretability of neuro-behavioral decoding, driving the field from black-box prediction toward a mechanism-driven scientific discovery platform.