A5071032466- Advanced Neural Network Applications
- Natural Language Processing Techniques
- Topic Modeling
- Multimodal Machine Learning Applications
- Advanced Image and Video Retrieval Techniques
- Remote-Sensing Image Classification
- RNA modifications and cancer
- RNA Research and Splicing
- Ferroptosis and cancer prognosis
- Video Surveillance and Tracking Methods
- Cancer, Hypoxia, and Metabolism
- Domain Adaptation and Few-Shot Learning
- Electronic and Structural Properties of Oxides
- Metabolism, Diabetes, and Cancer
- RNA and protein synthesis mechanisms
- Ubiquitin and proteasome pathways
- Cancer-related Molecular Pathways
- Remote Sensing and LiDAR Applications
- Automated Road and Building Extraction
- Histone Deacetylase Inhibitors Research
- Colorectal Cancer Treatments and Studies
- Magnetic and transport properties of perovskites and related materials
- Advanced Materials Characterization Techniques
- Cancer Mechanisms and Therapy
- Plant Gene Expression Analysis
Zhejiang Chinese Medical University
2024-2025
Guangdong Provincial People's Hospital
2022-2025
Guangdong Academy of Medical Sciences
2022-2025
Southern Medical University
2023-2025
Hangzhou First People's Hospital
2024-2025
Megvii (China)
2024
Westlake University
2024
Chinese Academy of Sciences
2020-2023
Aerospace Information Research Institute
2020-2023
University of Chinese Academy of Sciences
2019-2023
Two hallmarks for cancer cells are the accelerated cell cycle progression as well altered metabolism, however, how these changes coordinated to optimize growth advantage still poorly understood. Here we identify that Polo-like kinase 1 (Plk1), a key regulator mitosis, plays critical role biosynthesis in through activating pentose phosphate pathway (PPP). We find Plk1 interacts with and directly phosphorylates glucose-6-phosphate dehydrogenase (G6PD). By G6PD promoting formation of its active...
Arbitrary-oriented object detection is an important task in the field of remote sensing detection. Existing studies have shown that polar coordinate system has obvious advantages dealing with problem rotating modeling, is, using fewer parameters to achieve more accurate However, present state-of-the-art detectors based on deep learning are all modeled Cartesian coordinates. In this article, we introduce detector for first time, and propose anchor free Polar Remote Sensing Object Detector...
Motivated by the development of deep convolution neural networks (DCNNs), aircraft detection has gained tremendous progress. State-of-the-art DCNN-based detectors mainly belong to top-down approaches, which enumerate massive potential locations with form rectangular regions, and then identify whether they are objects or not. Compared these detectors, this article shows that via a type bottom-up method can have better performances in era learning. In article, we propose novel detector named...
We present the design, implementation and engineering experience in building deploying MegaScale, a production system for training large language models (LLMs) at scale of more than 10,000 GPUs. Training LLMs this brings unprecedented challenges to efficiency stability. take full-stack approach that co-designs algorithmic components across model block optimizer computation communication overlapping, operator optimization, data pipeline, network performance tuning. Maintaining high throughout...
Abstract DIS3-like 3′-5′ exoribonuclease 2 (DIS3L2) degrades aberrant RNAs, however, its function in tumorigenesis remains largely unexplored. Here, DIS3L2 expression promoted human hepatocellular carcinoma (HCC) progression via heterogeneous nuclear ribonucleoproteins (hnRNP) U-mediated alternative splicing. directly interacted with hnRNP U through cold-shock domains and inclusion of exon 3b during splicing pre-Rac1 independent exonuclease activity, yielding an oncogenic variant, Rac1b,...
Histone deacetylases (HDACs) engage in the regulation of various cellular processes by controlling global gene expression. The dysregulation HDACs leads to carcinogenesis, making ideal targets for cancer therapy. However, use HDAC inhibitors (HDACi) as single agents has been shown have limited success treating solid tumors clinical studies. This study aimed identify a novel downstream effector provide potential target combination therapy.Transcriptome sequencing and bioinformatics analysis...
While large scale pre-trained language models such as BERT have achieved great success on various natural understanding tasks, how to efficiently and effectively incorporate them into sequence-to-sequence the corresponding text generation tasks remains a non-trivial problem. In this paper, we propose address problem by taking two different encoder decoder respectively, fine-tuning introducing simple lightweight adapter modules, which are inserted between layers tuned task-specific dataset....
Hao-Ran Wei, Shujian Huang, Ran Wang, Xin-yu Dai, Jiajun Chen. Proceedings of the 2019 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019.
Arbitrary-oriented object detection is an important task in the field of remote sensing detection. Existing studies have shown that polar coordinate system has obvious advantages dealing with problem rotating modeling, is, using fewer parameters to achieve more accurate However, present state-of-the-art detectors based on deep learning are all modeled Cartesian coordinates. In this article, we introduce detector for first time, and propose anchor free Polar Remote Sensing Object Detector...
Transformer-based detector is a new paradigm in object detection, which aims to achieve pretty-well performance while eliminates the priori knowledge driven components, e.g., anchors, proposals and NMS. DETR, state-of-the-art model among them, composed of three sub-modules, i.e., CNN-based backbone paired transformer encoder-decoder. The CNN applied extract local features used capture global contexts. This pipeline, however, not concise enough. In this paper, we propose WB-DETR (DETR-based...
Enolase 1 (ENO1) is a glycolytic enzyme involved in tumour progression that performs variety of classical and nonclassical functions. However, the mechanism by which it promotes still not fully understood. Here, we found ENO1 can bind to β-site amyloid precursor protein cleaving 2 (BACE2), codependent gene ENO1, liver cancer cells. By suppressing lysosomal-dependent degradation, stabilises BACE2 level without affecting its mRNA level. Further analysis revealed promote low-density lipoprotein...
In the pursuit of superior video-processing MLLMs, we have encountered a perplexing paradox: "anti-scaling law", where more data and larger models lead to worse performance. This study unmasks culprit: "temporal hacking", phenomenon shortcut by fixating on select frames, missing full video narrative. this work, systematically establish comprehensive theory temporal hacking, defining it from reinforcement learning perspective, introducing Temporal Perplexity (TPL) score assess misalignment,...
Staphylococcal nuclease domain-containing protein 1 (SND1) is an evolutionarily conserved multifunctional that functions mainly in the nucleus and cytoplasm. However, whether SND1 regulates cellular activity through mitochondrial-related remains unclear. Herein, we demonstrate localized to mitochondria promote phosphoglycerate mutase 5 (PGAM5)-mediated mitophagy. We find present based on mass spectrometry data verified this phenomenon different liver cancer cell types by performing organelle...
Transformer model (Vaswani et al. 2017) has been widely used in machine translation tasks and obtained state-of-the-art results. In this paper, we report an interesting phenomenon its encoder-decoder multi-head attention: different attention heads of the final decoder layer align to word candidates. We empirically verify discovery propose a method generate diverse translations by manipulating heads. Furthermore, make use these with back-translation technique for better data augmentation....
Remote sensing images semantic segmentation is a fundamental yet challenging task, which has long relied heavily on sufficient pixelwise annotations. Semisupervised learning proposed to address the problem of high dependence labeled data by exploiting more learnable samples generated from large amounts accessible unlabeled data. However, affected complexity and diversity remote images, various misclassifications often occur lead errors accumulation during model training. Errors will destroy...
Abstract Background Latexin, also known as endogenous carboxypeptidase inhibitor (CPI), has been found to inhibit mouse stem cell populations and lymphoma proliferation, demonstrating its potential role a tumor suppressor. Our previous study suggested correlation between latexin expression malignant transformation of immortalized human gastric epithelial cells. Here, we examined in carcinomas investigated the effect differential on proliferation cancer cells vitro vivo . Methods Monoclonal...
Object detectors of remote sensing (RS) imagery with deep learning have become increasingly popular and rely heavily on extensive labeled data. The source-only detectors, which are trained massive data in a source domain, some cases fail to get satisfactory performance target domain due the shift. To alleviate shift, approaches consider feature distribution alignment, but unlabeled is under-utilized. Some methods use model generate pseudo labels for data, variation different scenarios...