A peptide bridge is designed to construct a fusion enzyme with electrostatic cofactor channeling, reducing NADPH input by two orders of magnitude or decreasing reaction time threefold with the same cofactor input.

Newly sequenced ancient genomes from Yunnan, China, have shed new light on human prehistory in East Asia. In a study published in Science, a research team led by Prof. FU Qiaomei at the Institute of Vertebrate Paleontology and Paleoanthropology of the Chinese Academy of Sciences analyzed data from 127 ancient humans, dating from 7,100 to 1,400 years ago. The results show that this region is pivotal to understanding the origin of both Tibetan and Austroasiatic (i.e., ethnic groups with a shared language group in South and Southeast Asia) population groups.

Current commercial UV-emitting materials rely heavily on non-sustainable resources such as rare metals, heavy metals, and petroleum-based chemicals. Recently, carbon dots have been synthesized from a renewable feedstock—green tea extract. These carbon dots exhibit UV emission in water. Interestingly, in poorer solvents, their emission blue-shifts and becomes nearly five times more efficient due to aggregation-induced emission behavior.

Inspired by the suckerfishes-shark motion behavior, they designed and prepared a kind of NIR light-propelled micro@nanomotor with weak acid-triggered release of H₂O₂-driven nanomotor. By the coordinated bond interaction, a large amount of Janus Au-Pt nanomotors with hydrogen peroxide (H₂O₂)-driven capacity, analogous to suckerfishes, were attached onto immovable yolk-shell structured polydopamine-mesoporous silica (PDA-MS) micromotor as the host to create two-stage PDA-MS@Au-Pt micro@nanomotor. PDA-MS@Au-Pt micro@nanomotor moved directionally by self-thermophoresis under the propulsion of NIR light with low power density. When the PDA-MS@Au-Pt entered into the weak acidic environment formed by a low concentration of H₂O₂, most small Au-Pt nanomotors were detached from the surface of PDA-MS due to the weak acidic sensitivity of the coordinated bond, and then performed self-diffusiophoresis in the environment containing a low concentration of H₂O₂ as a chemical fuel.

A groundbreaking non-hand-worn VR hand rehabilitation system has been developed, utilizing ionic hydrogel electrodes and deep learning for electromyography (EMG) gesture recognition. The system offers load-free rehabilitation without bulky mechanical components, providing a more accessible and flexible alternative to traditional rehabilitation methods. This VR-based solution enables immersive training and precise hand rehabilitation for stroke and joint disease patients in the comfort of their homes, without the constraints of time or location.

A self-supervised deep learning model has been developed to improve the quality of dynamic fluorescence images by leveraging temporal gradients. The method enables accurate denoising without ground truth data and facilitates clearer visualization of spatially and temporally dynamic biological signals in vivo.

Researchers from University of South China, Tsinghua University and Technical University of Munich have developed a whole system uncertainty model and an Intelligent optimized power control system of the space nuclear reactor with faster response, higher control accuracy and stronger adaptability under uncertainty conditions. These research results provide new ideas and solutions for improving the intelligence level and autonomous control capability of advanced nuclear energy systems in complex environments.

Understanding how cities grow is vital for shaping sustainable urban futures—but mapping the true extent of urban expansion remains a formidable technical hurdle. 

A research paper by scientists at Chinese Academy of Sciences proposed a dual-task learning framework, the “Twin Brother” model, which fuses convolutional neural network (CNN), long short-term memory (LSTM), neural networks (NNs), and the squeezing-elicited attention mechanism to classify the lateral gait stage and estimate the hip angle from electromyography (EMG) signals.

The new research paper, published on May. 1 in the journal Cyborg and Bionic Systems, provide a “Twin Brother” model. The model is a dual-task learning framework designed for simultaneous gait phases recognition for lateral walking and continuous hip angle prediction.