Article
Computer Science, Artificial Intelligence
Bin Liu, Konstantinos Blekas, Grigorios Tsoumakas
Summary: Class imbalance is a common challenge in multi-label data, and sampling techniques can be an effective strategy to address it. The imbalance level within the local neighborhood of minority class examples is crucial for performance degradation. The proposed sampling approaches, MLSOL and MLUL, are effective in alleviating the local label imbalance and improving performance on multi-label datasets.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Information Systems
Xiaoyan Zhu, Jiaxuan Li, Jingtao Ren, Jiayin Wang, Guangtao Wang
Summary: This study proposes a new method called MLDE for solving the multi-label classification problem. It selects the most competent ensemble of base classifiers to predict each unseen instance, effectively utilizing label correlation and achieving better performance.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Ran Wang, Sam Kwong, Xu Wang, Yuheng Jia
Summary: RAkEL is a multi-label learning strategy that integrates many single-label learning models, with randomly generated label subsets potentially leading to difficulties in separability and balance. The proposed ACkEL paradigm aims to address these issues by introducing an active label-selection criterion.
PATTERN RECOGNITION
(2021)
Article
Computer Science, Artificial Intelligence
Qingshuo Zhang, Eric C. C. Tsang, Qiang He, Yanting Guo
Summary: Multi-label learning is a type of machine learning that addresses the classification of data with multiple labels. Ensemble-based methods are commonly used in multi-label learning, but they typically use binary or multi-class classifiers as base learners. This study proposes an efficient multi-label classification method based on kernel extreme learning machine and ensemble learning, that addresses the time complexity and performance issues associated with existing methods. The experimental results demonstrate the superiority of the proposed method compared to baseline methods and other ensemble-based multi-label methods.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Automation & Control Systems
Yi-Bo Wang, Jun-Yi Hang, Min-Ling Zhang
Summary: Multi-label learning aims to assign a set of relevant class labels for an unseen instance by extracting label-specific features.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2022)
Article
Computer Science, Information Systems
Yaojin Lin, Qinghua Hu, Jinghua Liu, Xingquan Zhu, Xindong Wu
Summary: This article proposes a multi-label learning algorithm MULFE, which combines multiple label-specific feature spaces, label correlations, and weighted ensemble principle to achieve the maximum margin multi-label classification goal. Experimental studies on 10 public datasets demonstrate the effectiveness of MULFE in achieving accurate multi-label classification.
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA
(2022)
Article
Mathematical & Computational Biology
Jinghang Gu, Emmanuele Chersoni, Xing Wang, Chu-Ren Huang, Longhua Qian, Guodong Zhou
Summary: This article presents the work on the BioCreative VII LitCovid Track, proposing the LitCovid Ensemble Learning (LCEL) method for extracting semantic topics from COVID-19 literature. By integrating multiple biomedical pretrained models and utilizing additional biomedical knowledge and data augmentation, the representation abilities and learning deficiency of deep neural models are improved. Furthermore, an asymmetric loss function and ensemble bagging strategy are applied to achieve state-of-the-art performance on the LitCovid dataset.
DATABASE-THE JOURNAL OF BIOLOGICAL DATABASES AND CURATION
(2022)
Article
Computer Science, Information Systems
Khudran M. Alzhrani
Summary: This paper discusses a novel ensemble strategy for multi-output neural networks in multi-label classification tasks, demonstrating that ensemble prediction from output layers can enhance network performance. By selecting baseline and proposed models based on hidden layer size and output layer number, the ENSOCOM method improved network performance across multiple datasets.
CMC-COMPUTERS MATERIALS & CONTINUA
(2022)
Article
Computer Science, Information Systems
Bo Liu, Weibin Li, Yanshan Xiao, Xiaodong Chen, Laiwang Liu, Changdong Liu, Kai Wang, Peng Sun
Summary: Multi-label learning is a popular topic in machine learning that deals with the simultaneous association of multiple labels with given samples. This paper proposes a new multi-view multi-label learning method called ELSMML, which considers label correlation. The method constructs a crafted label correlation matrix to describe label relationships and utilizes multi-view learning and dimension reduction to exploit latent semantic label information and feature information, building a classifier in a low dimensional space. The ELSMML model is optimized using the accelerated proximal gradient method and achieves better performance compared to other baselines according to evaluation metrics.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Bin -Bin Jia, Min -Ling Zhang
Summary: In this paper, a new classification framework called Multi-Dimensional Multi-Label (MDML) classification is investigated, which models objects with rich semantics by encompassing heterogeneous label spaces and multi-label annotations. A learning method named CLIM is proposed to learn from MDML training examples. CLIM induces base multi-label predictive models w.r.t. each label space and uses thresholding predictions to augment the original feature space and yield stacked multi-label predictive models. Experiments on real-world MDML data sets validate the effectiveness of CLIM.
PATTERN RECOGNITION
(2023)
Article
Pharmacology & Pharmacy
Haozheng Li, Yihe Pang, Bin Liu, Liang Yu
Summary: This study highlights the importance and functions of intrinsically disordered regions (IDRs) and their molecular recognition features (MoRFs) in protein structures. Understanding the functions of MoRFs is crucial for pharmaceutical and disease pathogenesis. However, existing computational methods fail to distinguish the different functions of MoRFs. In this paper, a multi-label learning approach using the Binary Relevance (BR) strategy and ensemble learning techniques is proposed to predict MoRF functions. The experimental results show that the MoRF-FUNCpred model performs well in predicting MoRF functions. To the best of our knowledge, MoRF-FUNCpred is the first predictor for predicting MoRF functions.
FRONTIERS IN PHARMACOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Abdelouahid Alalga, Khalid Benabdeslem, Dou El Kefel Mansouri
Summary: This paper introduces a new approach in semi-supervised multi-label learning and proposes an ensemble method in feature selection to enhance the stability and performance of the algorithm. Experimental results demonstrate the superior performance of this method in the classification task.
KNOWLEDGE AND INFORMATION SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Lijuan Sun, Songhe Feng, Gengyu Lyu, Hua Zhang, Guojun Dai
Summary: This article introduces a novel approach to partial multi-label learning, which deals with noisy side information by removing noisy outliers from the training instances and training robust partial multi-label classifier for unlabeled instances prediction.
KNOWLEDGE AND INFORMATION SYSTEMS
(2021)
Article
Biochemical Research Methods
Yushuang Liu, Shuping Jin, Hongli Gao, Xue Wang, Congjing Wang, Weifeng Zhou, Bin Yu
Summary: This article proposes a novel method called ML-locMLFE, which can effectively predict the multi-label subcellular localization of proteins and has obvious advantages. By using different feature extraction methods and information processing methods, this method demonstrates good accuracy in predicting the protein localization of diseases such as SARS-CoV-2.
Article
Computer Science, Information Systems
Liang Zhao, Yanshan Xiao, Bo Liu, Zhifeng Hao
Summary: "PLL is a method for handling partial label learning, aiming to identify the ground-truth label from a set of candidate labels. Most existing PLL approaches focus on the single-view problem and do not address the multi-view problem. This paper proposes a novel multi-view paunknown method."
INFORMATION SCIENCES
(2022)
Article
Biochemical Research Methods
Feng Huang, Xiang Yue, Zhankun Xiong, Zhouxin Yu, Shichao Liu, Wen Zhang
Summary: MicroRNAs are crucial in human diseases, and studying their associations can help understand the molecular mechanisms of these diseases. This study introduces a tensor decomposition method to predict multi-type miRNA-disease associations, improving current baseline performance and incorporating biological features for better efficiency.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Biochemical Research Methods
Zhouxin Yu, Feng Huang, Xiaohan Zhao, Wenjie Xiao, Wen Zhang
Summary: The paper introduces a novel computational method LAGCN for predicting drug-disease associations, which outperforms existing state-of-the-art methods. By integrating embeddings from multiple graph convolution layers and combining them using the attention mechanism, LAGCN achieves efficient drug-disease association prediction.
BRIEFINGS IN BIOINFORMATICS
(2021)
Review
Biotechnology & Applied Microbiology
Xinran Xu, Shuai Liu, Zhihao Yang, Xiaohan Zhao, Yaozhen Deng, Guangzhan Zhang, Jian Pang, Chengshuai Zhao, Wen Zhang
Summary: This review introduces the development of computational methods for lncRNA prediction and presents a new Python package, ezLncPred, which provides a convenient way to utilize nine state-of-the-art lncRNA prediction methods. The challenges and future directions of lncRNA prediction are also discussed in the paper.
BRIEFINGS IN FUNCTIONAL GENOMICS
(2021)
Article
Biochemical Research Methods
Haitao Fu, Feng Huang, Xuan Liu, Yang Qiu, Wen Zhang
Summary: Motivation: There are various interaction/association bipartite networks in biomolecular systems, and identifying unobserved links in biomedical bipartite networks is crucial for understanding complex diseases. This study proposes a novel multi-view graph convolution network (MVGCN) framework for link prediction in biomedical bipartite networks by combining similarity networks and performing self-supervised learning. Results show that MVGCN outperforms baseline methods and has good generalization capacity on benchmark datasets.
Article
Biochemical Research Methods
Menglu Li, Wen Zhang
Summary: The article proposes a predictive method for phage-host interactions using a generative adversarial network and sequence-based feature fusion. The method outperforms other state-of-the-art prediction methods and is validated through a case study.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Guangzhan Zhang, Menglu Li, Huan Deng, Xinran Xu, Xuan Liu, Wen Zhang
Summary: MiRNAs, small non-coding RNA molecules, play a significant role in biological processes. This paper proposes a signed graph neural network method (SGNNMD) to predict the deregulation types of miRNA-disease associations, which have potential applications in drug development and clinical diagnosis. Experimental results demonstrate the competitive performance of SGNNMD.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Xuan Liu, Congzhi Song, Feng Huang, Haitao Fu, Wenjie Xiao, Wen Zhang
Summary: The paper proposed a graph neural network method with contrastive learning for predicting cancer cell line response to therapeutic drugs. The method outperformed existing methods in computational experiments and highlighted the importance of biological features, known responses, and contrastive learning for accurate prediction.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Computer Science, Interdisciplinary Applications
Hua Chen, Sheng Sheng, Chong-Yu Xu, Zhiyu Li, Wen Zhang, Shaowen Wang, Shenglian Guo
Summary: The spatiotemporal estimation method based on F-SVD improves rainfall estimations compared to IDW and OK, especially in cases with larger amounts of rainfall events. Combining with FSVD significantly enhances the accuracy of IDW and OK.
ENVIRONMENTAL MODELLING & SOFTWARE
(2021)
Review
Biochemical Research Methods
Yang Qiu, Yang Zhang, Yifan Deng, Shichao Liu, Wen Zhang
Summary: This paper provides a comprehensive review of computational methods for detecting drug-drug interactions (DDIs). It discusses three categories of methods: literature-based extraction methods, machine learning-based prediction methods, and pharmacovigilance-based data mining methods. The paper presents the research background, data sources, representative approaches, and evaluation metrics for each category. It also discusses the current challenges and potential opportunities for future directions in DDI detection.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Zhankun Xiong, Feng Huang, Ziyan Wang, Shichao Liu, Wen Zhang
Summary: Drug repositioning is important for finding new treatments for diseases, but efficiently utilizing biological datasets is challenging. This paper presents a multimodal framework called GraphPK for improving drug repositioning through utilizing prior knowledge from a drug knowledge graph. Experimental results show that the method outperforms other approaches and using prior knowledge from knowledge graphs enhances the model's predictive power and robustness.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Shichao Liu, Yang Zhang, Yuxin Cui, Yang Qiu, Yifan Deng, Zhongfei Zhang, Wen Zhang
Summary: Drug-drug interactions are a significant concern in drug discovery, and accurate prediction of these interactions is crucial for improving research efficiency and safety. This paper proposes a Deep Attention Neural Network based Drug-Drug Interaction prediction framework (DANN-DDI) that can predict unobserved drug-drug interactions. The framework utilizes multiple data sources and employs a deep neural network to accurately predict these interactions. Experimental results demonstrate the improved prediction performance of DANN-DDI compared to state-of-the-art methods, and its ability to predict novel drug-drug interactions and associated events.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Shen Han, Haitao Fu, Yuyang Wu, Ganglan Zhao, Zhenyu Song, Feng Huang, Zhongfei Zhang, Shichao Liu, Wen Zhang
Summary: Accurate prediction of molecular properties is important in drug discovery. Previous works have developed various representation schemes for capturing chemical information in molecules. This study proposes a novel framework, HimGNN, which combines atom- and motif-based graphs to learn hierarchical molecular topology representations. HimGNN achieves promising performances on classification and regression tasks in molecular property prediction.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Xuan Liu, Wen Zhang
Summary: SubCDR is a deep learning method for interpretable cancer drug response (CDR) prediction, which can identify the key subcomponents driving the response outcomes. By extracting functional subcomponents from drug and cell line profiles, and identifying the interactions between subcomponents, SubCDR can find the subcomponents that contribute more to the response outcomes. Extensive computational experiments on the GDSC dataset have demonstrated the superiority of SubCDR over existing methods, and many predicted cases have shown the potential of SubCDR in finding key subcomponents driving responses and discovering new therapeutic drugs. Rating: 8/10.
PLOS COMPUTATIONAL BIOLOGY
(2023)
Article
Biochemical Research Methods
Shuai Liu, Xinran Xu, Zhihao Yang, Xiaohan Zhao, Shichao Liu, Wen Zhang
Summary: In this paper, a deep neural network-based method named EPIHC is proposed to predict Enhancer-Promoter Interactions. EPIHC extracts features using convolutional neural networks and captures the communicative information between enhancer and promoter sequences through a communicative learning module. Experimental results show that EPIHC outperforms existing methods in EPI prediction and the communicative learning module provides explainability about EPIs.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Xian-gan Chen, Shuai Liu, Wen Zhang
Summary: Non-coding RNAs are important in biological processes and disease association. Predicting coding potential of RNA sequences is crucial for downstream analysis, but current methods perform poorly on short Open Reading Frame (sORF) RNA sequences. This study presents a coding potential prediction method, CPE-SLDI, which uses data oversampling to address the problem of local data imbalance in sORF sequences, resulting in improved performance.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2022)