Article
Microbiology
Yu Peng, Shouwei Zhao, Zhiliang Zeng, Xiang Hu, Zhixiang Yin
Summary: Prediction of drug-target interactions (DTIs) is crucial for drug development. Traditional laboratory methods are time-consuming and expensive. Recent studies have shown that using machine learning methods, such as LightGBM, can accelerate the drug development process and reduce costs. In this study, we compared the performance of our new model, LGBMDF, with other state-of-the-art methods using cross-validation, and found that our method has better accuracy and faster computation.
FRONTIERS IN MICROBIOLOGY
(2023)
Article
Genetics & Heredity
Ying Zheng, Zheng Wu
Summary: Drug repositioning is an effective method for predicting drug-target interactions, utilizing heterogeneous networks to construct similarity matrix and employing cascade deep forest method for prediction.
FRONTIERS IN GENETICS
(2021)
Article
Computer Science, Artificial Intelligence
Junzhong Ji, Junwei Li
Summary: Deep forest is a new multi-layer ensemble model, but its large-scale application is hindered by high time costs and storage requirements. To address these issues, we propose a tri-objective optimization-based cascade ensemble pruning (TOOCEP) algorithm. We first present a tri-objective optimization-based single-layer pruning (TOOSLP) method to prune the single-layer of Deep forest, considering accuracy, independent diversity, and coupled diversity. Then, we perform TOOSLP in a cascade framework to prune Deep forest layer-by-layer. Experimental results show that TOOCEP outperforms state-of-the-art methods in accuracy and pruned rate, reducing storage space and speeding up prediction.
PATTERN RECOGNITION
(2023)
Article
Biochemical Research Methods
Jilong Bian, Xi Zhang, Xiying Zhang, Dali Xu, Guohua Wang
Summary: Accurate and effective drug-target interaction (DTI) prediction is crucial in speeding up drug development and reducing costs. This study proposes a shared-weight-based MultiheadCrossAttention mechanism to improve the accuracy and efficiency of DTI prediction. Experimental results on six public drug-target datasets demonstrate the superiority of the proposed method over existing baselines.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biochemistry & Molecular Biology
Aida Tayebi, Niloofar Yousefi, Mehdi Yazdani-Jahromi, Elayaraja Kolanthai, Craig J. Neal, Sudipta Seal, Ozlem Ozmen Garibay
Summary: This study proposes a computational framework for predicting drug-target interaction using an ensemble of deep learning models to address the class imbalance problem in DTI prediction. The results from computational and experimental validations demonstrate the superiority of the proposed model over unbalanced models, highlighting the importance of balancing in improving performance.
Review
Computer Science, Information Systems
Ammar Mohammed, Rania Kora
Summary: Ensemble learning and deep learning are two approaches in machine learning that outperform traditional algorithms. Ensemble learning integrates multiple base models to create a stronger model, while deep learning-based models improve predictive accuracy across various domains. Recent research efforts have focused on combining ensemble learning with deep learning to overcome the challenge of tuning optimal hyperparameters.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Arjun Puri, Manoj Kumar Gupta, Kanica Sachdev
Summary: This article proposes a model for studying drug-target interaction problems using computational techniques. The model uses feature representations and resampling techniques to handle class imbalance, and utilizes a soft voting ensemble method to improve prediction accuracy. Experiments demonstrate that the proposed model outperforms existing models on standard datasets.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Biochemical Research Methods
Weiping Lin, Lianlian Wu, Yixin Zhang, Yuqi Wen, Bowei Yan, Chong Dai, Kunhong Liu, Song He, Xiaochen Bo
Summary: Combination therapy has shown effective results in treating complex diseases, but the search space for drug combinations is too large to be experimentally validated. Artificial intelligence techniques, particularly machine learning methods, have been used to discover synergistic drug combinations and significantly reduce experimental workload. In this study, a novel approach called EC-DFR is presented to predict synergistic drug combinations in various cancer cell lines by incorporating cell line-specific drug-induced gene expression profiles as features. The EC-DFR outperforms other methods and has been validated through biological experiments. The analysis also highlights the importance of cellular responses of drugs in predicting synergism.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Jiande Huang, Ping Chen, Lijuan Lu, Yuhui Deng, Qiang Zou
Summary: The paper proposes a Weighted Cascade Deep Forest framework (WCDForest) that addresses overfitting and characteristic dispersion issues in the deep forest model. The framework uses a multi-grained scanning module and a class vector weighting module to enhance performance, and introduces a feature enhancement module to reduce information loss. Experimental results demonstrate that WCDForest outperforms existing models.
APPLIED INTELLIGENCE
(2023)
Article
Mathematical & Computational Biology
Sofia D'Souza, K. V. Prema, S. Balaji, Ronak Shah
Summary: Chemogenomics, or proteochemometrics, uses computational methods to predict drug-target interactions based on large-scale data. This study develops a deep learning CNN model using one-dimensional SMILES for drugs and protein binding pocket sequences as inputs to predict unknown ligand-target interactions. The proposed method outperforms shallow machine learning methods in terms of prediction accuracy and computational efficiency.
INTERDISCIPLINARY SCIENCES-COMPUTATIONAL LIFE SCIENCES
(2023)
Review
Biochemistry & Molecular Biology
Karim Abbasi, Parvin Razzaghi, Antti Poso, Saber Ghanbari-Ara, Ali Masoudi-Nejad
Summary: This study provides a comprehensive overview of deep learning-based drug-target interactions (DTIs) prediction approaches, exploring different deep network architectures, commonly used datasets, and current challenges in the field. The research findings reveal the potential of deep learning in DTIs prediction and suggest future directions for development.
CURRENT MEDICINAL CHEMISTRY
(2021)
Article
Biochemical Research Methods
Siyuan Liu, Yusong Wang, Yifan Deng, Liang He, Bin Shao, Jian Yin, Nanning Zheng, Tie-Yan Liu, Tong Wang
Summary: This study proposes a novel approach called IGT that improves the prediction performance of active binding drugs in virtual screening. Compared to existing methods, IGT achieves better results in binding activity and binding pose prediction, and demonstrates superior generalization ability to unseen receptor proteins. Furthermore, IGT shows promising accuracy in drug screening against severe acute respiratory syndrome coronavirus 2.
BRIEFINGS IN BIOINFORMATICS
(2022)
Review
Biochemistry & Molecular Biology
Azwaar Khan Azlim Khan, Nurul Hashimah Ahamed Hassain Malim
Summary: The prediction of drug-target interactions (DTIs) is crucial in drug discovery, and machine learning and deep learning methods have been successful in accurately predicting DTIs. However, the imbalanced and high-dimensional nature of the datasets used poses a challenge, which can be addressed by resampling techniques. This paper compares different data resampling techniques and evaluates the effectiveness of deep learning methods in overcoming class imbalance in predicting DTIs.
Review
Biochemical Research Methods
Yuni Zeng, Xiangru Chen, Yujie Luo, Xuedong Li, Dezhong Peng
Summary: In this study, an end-to-end model with multiple attention blocks was proposed to predict the binding affinity scores of drug-target pairs. The model encodes correlations between atoms using a relation-aware self-attention block and models the interaction between drug and target representations using a multi-head attention block. Experimental results show that the proposed approach outperforms existing methods by benefiting from encoded correlation and extracted interaction information.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Orel Lavie, Asaf Shabtai, Gilad Katz
Summary: This study proposes methods for fine-tuning and calibrating DRL-based policies to meet multiple performance goals. It also presents a method for transferring effective security policies from one dataset to another. Furthermore, it demonstrates the robustness of the approach against adversarial attacks.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Biochemical Research Methods
Aman Chandra Kaushik, Yan-Jing Wang, Xiangeng Wang, Dong-Qing Wei
Summary: The study revealed significant differences in the genomics, transcriptomics, methylomics, and molecular dynamics between KRAS and TP53 mutations from the wild type in PAAD, with the prognosis of pancreatic cancer directly linked to specific KRAS mutations and protein stability. Screened drugs show potential effectiveness for PAAD patients.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Biochemical Research Methods
Hao Wu, Yingfu Wu, Yuhong Jiang, Bing Zhou, Haoru Zhou, Zhongli Chen, Yi Xiong, Quanzhong Liu, Hongming Zhang
Summary: This study proposes a high accuracy cell classification algorithm, scHiCStackL, based on single-cell Hi-C data. The algorithm improves the data preprocessing method and constructs a two-layer stacking ensemble model for classifying cells. Experimental results show that scHiCStackL achieves superior performance in predicting cell types using single-cell Hi-C data.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Shenggeng Lin, Yanjing Wang, Lingfeng Zhang, Yanyi Chu, Yatong Liu, Yitian Fang, Mingming Jiang, Qiankun Wang, Bowen Zhao, Yi Xiong, Dong-Qing Wei
Summary: One of the main problems with the joint use of multiple drugs is the potential for adverse drug interactions and side effects. This study proposes a novel method, MDF-SA-DDI, which predicts drug-drug interaction events using multi-source drug fusion, multi-source feature fusion, and transformer self-attention mechanism.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Yumeng Zhang, Yangming Zhang, Yi Xiong, Hui Wang, Zixin Deng, Jiangning Song, Hong-Yu Ou
Summary: Bacterial type IV secretion systems (T4SSs) play crucial roles in bacterial pathogenesis, with versatile functions. Using a sequence embedding strategy from a pre-trained language model can enhance the accuracy and speed of T4SE prediction tools.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Jianmin Wang, Yanyi Chu, Jiashun Mao, Hyeon-Nae Jeon, Haiyan Jin, Amir Zeb, Yuil Jang, Kwang-Hwi Cho, Tao Song, Kyoung Tai No
Summary: This study constructs a dataset for protein-protein interaction (PPI) targeted drug-likeness and proposes a deep molecular generative framework to generate novel drug-like molecules based on the features of seed compounds. The results show that the generated molecules have better PPI-targeted drug-likeness and drug-likeness, and the model performs comparably to other state-of-the-art molecule generation models.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biology
Shenggeng Lin, Guangwei Zhang, Dong-Qing Wei, Yi Xiong
Summary: Polypharmacy is an effective strategy for treating complex or co-existing diseases, but it can lead to higher risk of adverse side effects due to drug interactions. This study proposes a deep learning-based method called DeepPSE to predict polypharmacy side effects. Experimental results demonstrate that DeepPSE outperforms five baseline or state-of-the-art methods in predicting drug pair side effects.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Biochemical Research Methods
Jing Zhao, Bowen Zhao, Xiaotong Song, Chujun Lyu, Weizhi Chen, Yi Xiong, Dong-Qing Wei
Summary: The Subtype-DCC method, which integrates multi-omics data, is proposed for cancer subtyping and demonstrates superior performance compared to existing clustering methods. It has potential applications in cancer diagnosis, prognosis, and treatment.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Zhiqiang Hu, Wenfeng Liu, Chenbin Zhang, Jiawen Huang, Shaoting Zhang, Huiqun Yu, Yi Xiong, Hao Liu, Song Ke, Liang Hong
Summary: Drug-target binding affinity prediction is a crucial task in drug discovery. Surprisingly, this study shows that a model can achieve superior performance without any protein sequence information. Instead, the model characterizes proteins based on the ligands they interact with. The proposed paradigm outperforms sequence-based methods in terms of Mean Squared Error (MSE) and R-Square.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Chemistry, Medicinal
Song Li, Chao Hu, Song Ke, Chenxing Yang, Jun Chen, Yi Xiong, Hao Liu, Liang Hong
Summary: This article presents a novel ligand and structure integrated molecular generative model called LS-MolGen, which combines representation learning, transfer learning, and reinforcement learning. The model demonstrates superior performance in generating promising compounds with novel scaffolds and high binding affinity for drug design.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Genetics & Heredity
Shaojun Wang, Ronghui You, Yunjia Liu, Yi Xiong, Shanfeng Zhu
Summary: NetGO 2.0 is an automated function prediction method that integrates multiple sources of information to improve performance. However, it does not leverage valuable information from a large number of unannotated proteins. Recently, protein language models have been proposed to learn informative representations from protein sequences. By incorporating this model into NetGO 2.0, researchers have developed NetGO 3.0, which significantly improves the performance of automated function prediction.
GENOMICS PROTEOMICS & BIOINFORMATICS
(2023)
Article
Chemistry, Medicinal
Heqi Sun, Jianmin Wang, Hongyan Wu, Shenggeng Lin, Junwei Chen, Jinghua Wei, Shuai Lv, Yi Xiong, Dong-Qing Wei
Summary: This study proposes a deep learning framework called MultiPPIMI for predicting the interaction between PPI targets and modulators. Experimental results show that MultiPPIMI performs well in predicting both cold-start scenarios and random-split scenarios, and can assist in screening inhibitors for Keap1/Nrf2 PPI interactions.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Biochemical Research Methods
Shenggeng Lin, Xueying Mao, Liang Hong, Shuangjun Lin, Dong-Qing Wei, Yi Xiong
Summary: This article introduces a novel method, MATT-DDI, for predicting multi-type drug-drug interactions (DDIs) by accurately considering the original feature vectors of drugs and multiple attention mechanisms. Experimental results show that MATT-DDI has good performance and no information leakage.
Article
Biochemical Research Methods
Zhiqiang Hu, Wenfeng Liu, Chenbin Zhang, Jiawen Huang, Shaoting Zhang, Huiqun Yu, Yi Xiong, Hao Liu, Song Ke, Liang Hong
Summary: This study demonstrates that drug-target binding affinity prediction can achieve superior performance without accessing protein-sequence-related information. Instead, the study characterizes a protein solely based on the ligands it interacts with. By separately treating different proteins and training them jointly, a robust and universal representation of ligands is learned, which outperforms sequence-based methods. The results highlight the model's robustness in generalizing to unseen proteins and predicting future data.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Yanyi Chu, Yan Zhang, Qiankun Wang, Lingfeng Zhang, Xuhong Wang, Yanjing Wang, Dennis Russell Salahub, Qin Xu, Jianmin Wang, Xue Jiang, Yi Xiong, Dong-Qing Wei
Summary: The study presents the TransMut framework, which automatically optimizes mutated peptides with higher affinity to the target HLA allele. This framework provides an automated approach for screening immunogenic peptides and vaccine design.
NATURE MACHINE INTELLIGENCE
(2022)