The team of Professor Heng-Yu Fan from the Life Sciences Institute (LSI) of Zhejiang University (with ph.D. student Yuxuan Jiao as the first author) published a research paper titled “DHX36-mediated G-quadruplexes unwinding is essential for oocyte and early embryo development in mice” in Science Bulletin. This study, for the first time, revealed the spatiotemporal consistency of DHX36 with the formation of G4 structures during mammalian oocyte and early embryo development. DHX36 specifically binds to G4 structures on DNA (particularly in rDNA regions and transcription start sites) and pre-rRNA, thereby regulating chromatin conformation, gene transcription, and pre-rRNA processing, which are essential for maintaining normal oocyte and embryo development.

This work presents an energy-efficient classical simulation algorithm using GPUs that outperforms Google's Sycamore processor in speed and energy consumption, challenging the claim of quantum computational advantage.

The paper advocates drilling intact oceanic crust to determine the Moho's petrological nature and resolve crustal composition misconceptions, highlighting China's new drilling vessel's role. 

A team of researchers from Peking University and the Chinese Academy of Sciences has made a breakthrough in quantum photonics by demonstrating a method to generate large-scale entangled states—known as cluster states—directly on a chip using optical microresonators. Their work, detailed in Light: Science & Applications, achieves a 60-mode cluster state, an order of magnitude larger than previous on-chip demonstrations.

A novel approach called Counterfactual Synthetic Minority Oversampling Technique (SMOTE) has been developed to tackle the persistent issue of imbalanced data in healthcare. Traditional models trained on imbalanced datasets often overlook rare but critical conditions, such as diseases, leading to biased predictions. By combining SMOTE with counterfactual generation, Counterfactual SMOTE creates synthetic data points near decision boundaries while minimizing noise. This method significantly enhances model performance, enabling better detection of rare conditions without overwhelming false positives. Tested across 24 healthcare datasets, Counterfactual SMOTE outperforms existing methods, offering a robust solution for improving medical diagnostics and extending to other fields.

A recent study introduces an advanced anomaly-based intrusion detection system (IDS) designed to address the increasing cyber threats targeting Internet of Things (IoT) devices. By combining machine learning (ML) and deep learning (DL) models with particle swarm optimization (PSO) for feature selection, the system achieves remarkable performance. Tested on the RT_IoT2022 dataset, the system's top performer, CatBoost, achieved an impressive 99.85% accuracy, making it one of the most accurate IDS solutions for IoT environments. This innovation offers a powerful tool to detect and classify complex attacks in real-time, enhancing cybersecurity for resource-constrained IoT networks.