Review
Biochemistry & Molecular Biology
Sha Zhu, Qifeng Bai, Lanqing Li, Tingyang Xu
Summary: Drug repositioning plays a significant role in drug development and machine learning methods can accelerate this process. This article focuses on the repurposing potential of type 2 diabetes mellitus drugs for various diseases.
COMPUTATIONAL AND STRUCTURAL BIOTECHNOLOGY JOURNAL
(2022)
Article
Biochemical Research Methods
Kuo Yang, Yuxia Yang, Shuyue Fan, Jianan Xia, Qiguang Zheng, Xin Dong, Jun Liu, Qiong Liu, Lei Lei, Yingying Zhang, Bing Li, Zhuye Gao, Runshun Zhang, Baoyan Liu, Zhong Wang, Xuezhong Zhou
Summary: Drug repositioning is an important method in drug development that identifies new indications of approved drugs through analysis of clinical and experimental data. This study proposes a drug repositioning framework called DRONet, which combines network embedding and ranking learning to utilize the effectiveness comparative relationships among drugs. Experimental results show that DRONet achieves higher prediction accuracy than existing methods and demonstrates potential for guiding clinical drug development.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Computer Science, Information Systems
Phu Pham, Loan T. T. Nguyen, Ngoc Thanh Nguyen, Robert Kozma, Bay Vo
Summary: The integration of deep learning and fuzzy learning is a promising research direction in data embedding. It helps improve the performance of latent feature representation learning and multiple recommendation problem fine-tuning. However, existing deep learning-based recommendation techniques face major challenges related to data uncertainty and noise.
INFORMATION SCIENCES
(2023)
Review
Pharmacology & Pharmacy
Jun-Lin Yu, Qing-Qing Dai, Guo-Bo Li
Summary: Drug repositioning is an attractive strategy for discovering new therapeutic uses for approved or investigational drugs. Deep learning has gained attention for its potential in target prediction and drug repositioning, improving efficiency and success rates.
DRUG DISCOVERY TODAY
(2022)
Review
Chemistry, Medicinal
Tao Song, Gan Wang, Mao Ding, Alfonso Rodriguez-Paton, Xun Wang, Shudong Wang
Summary: This review presents a series of important network-based methods applied in drug repositioning, compares their development process, and highlights their significance in the field.
MOLECULAR INFORMATICS
(2022)
Article
Biochemical Research Methods
Bo-Wei Zhao, Xiao-Rui Su, Peng-Wei Hu, Yu-Peng Ma, Xi Zhou, Lun Hu
Summary: Drug repositioning is a strategy that uses artificial intelligence techniques to discover new indicators for approved drugs and improve traditional drug discovery and development. However, most computational methods fail to consider the non-Euclidean nature of biomedical network data. To address this, a deep learning framework called DDAGDL is proposed to predict drug-drug associations. Experimental results show that this method outperforms state-of-the-art drug repositioning methods in terms of several evaluation metrics.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Computer Science, Information Systems
Haodong Zou, Zhen Duan, Xinru Guo, Shu Zhao, Jie Chen, Yanping Zhang, Jie Tang
Summary: SANE extracts node sequence, attribute sequence, and label sequence in attributed networks, providing different insights into networks. By extracting bidirectional features from attribute sequence and using them to decode node and label sequences, SANE can scale to large networks and alleviate over-smoothing caused by attribute-only aggregation.
INFORMATION SCIENCES
(2021)
Article
Biochemical Research Methods
Yifan Shang, Xiucai Ye, Yasunori Futamura, Liang Yu, Tetsuya Sakurai
Summary: This study introduces a new computational framework, MccDTI, for predicting potential drug-target interactions using multiview network embedding, which outperforms other methods in prediction accuracy based on experimental results.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Automation & Control Systems
Benny Avelin, Anders Karlsson
Summary: We explore the dynamical and geometrical aspects of deep learning. By quantifying differences between data or decision functions, we present semi-invariant metrics that are applicable to many standard choices of layer maps. By considering random layer maps and employing non-commutative ergodic theorems, we are able to deduce the existence of certain limits as the number of layers tends to infinity. Additionally, we investigate the random initialization of standard networks and discover a surprising cutoff phenomenon in terms of the depth of the network. This cutoff phenomenon may be an important parameter for choosing an appropriate number of layers or a good initialization procedure. In conclusion, we hope that the concepts and results in this paper can provide a geometric framework for the theoretical understanding of deep neural networks.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Computer Science, Information Systems
Gopal Behera, Neeta Nain
Summary: The tremendous growth in information has led to overwhelming problems in accessing personalized products. To address these issues, we propose an efficient deep collaborative recommender system that embeds item metadata. This system utilizes neural networks and matrix factorization to handle the nonlinearity and sparsity of the data.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Karrar Al-Kaabi, Reza Monsefi, Davood Zabihzadeh
Summary: This paper presents a framework to enhance the generalization ability of existing deep metric learning methods in zero-shot learning tasks. By employing general yet discriminative representation learning and a class adversarial neural network, the framework addresses the limitations of DML methods in certain applications.
APPLIED INTELLIGENCE
(2023)
Article
Automation & Control Systems
Yangyang Li, Chaoqun Fei, Chuanqing Wang, Hongming Shan, Ruqian Lu
Summary: Deep metric learning methods integrate conventional metric learning with deep neural networks seamlessly, achieving great results on visual understanding tasks. Existing methods fail to preserve the geometric structure of data in the embedding space, leading to a shift in spatial structure and slow convergence. To address this, we propose the deep Riemannian metric learning (DRML) framework, which exploits the non-Euclidean geometric structural information. DRML outperforms existing methods on benchmark datasets, demonstrating its effectiveness.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2023)
Review
Biochemical Research Methods
Fei Wang, Yulian Ding, Xiujuan Lei, Bo Liao, Fang-Xiang Wu
Summary: Drug repositioning is an important method for exploring new uses of existing drugs in drug discovery, especially in pre-clinical stages. Computational approaches, including machine learning and deep learning, have shown great potential in saving time and reducing costs compared to traditional drug discovery methods.
CURRENT BIOINFORMATICS
(2022)
Article
Computer Science, Software Engineering
Arniel Labrada, Benjamin Bustos, Ivan Sipiran
Summary: Significant advances have been made in tasks like 3D model retrieval, classification, and segmentation. However, traditional 3D representations have limitations in cognitive processes due to their high redundancy and complexity. To address this, we propose a deep learning architecture that utilizes image views to represent 3D models, achieving high effectiveness in similarity assessment and improving training and inference times compared to state-of-the-art techniques.
Article
Chemistry, Analytical
Jingyi Liu, Caijuan Shi, Dongjing Tu, Ze Shi, Yazhi Liu
Summary: This paper proposes a zero-shot image classification method based on deep learning, which reduces the dependence on labeled training samples by using common space embedding and end-to-end learnable deep metric to learn the similarity of visual features and semantic features.