Article
Biochemical Research Methods
Chenran Wang, Yang Chen, Yuan Zhang, Keqiao Li, Menghan Lin, Feng Pan, Wei Wu, Jinfeng Zhang
Summary: Protein ligand docking is a computational tool for predicting protein functions and screening drug candidates. In this study, a novel reinforcement learning approach called A3C was developed to address the challenging problem of protein ligand docking. The experimental results showed significant improvement in binding site prediction compared to a naive model.
BMC BIOINFORMATICS
(2022)
Review
Biochemistry & Molecular Biology
Yuxiao Wang, Qihong Jiao, Jingxuan Wang, Xiaojun Cai, Wei Zhao, Xuefeng Cui
Summary: Predicting the binding affinities between target proteins and small molecule drugs is crucial for drug research and design. Deep learning methods have been widely utilized for precise affinity prediction. This study analyzes and discusses various deep learning methods, and conducts experiments to evaluate their prediction capabilities. By combining the strengths of the four models, improvements in prediction accuracy are achieved.
COMPUTATIONAL AND STRUCTURAL BIOTECHNOLOGY JOURNAL
(2023)
Article
Multidisciplinary Sciences
Zhenyu Meng, Kelin Xia
Summary: Molecular descriptors are crucial for quantitative structure-activity relationship (QSAR) models and machine learning-based data analysis. The proposed PerSpect ML models utilize a novel filtration process to generate spectral models at various scales, showing potential to greatly improve learning models in molecular data analysis. Results demonstrate superior performance in protein-ligand binding affinity prediction compared to existing models across commonly used databases.
Article
Chemistry, Medicinal
Xiaochen Bo, Song He, Shengqi Wang, Qingyu Li, Xiaochang Zhang, Lianlian Wu
Summary: In this study, a novel model named PLAMoRe was proposed to predict protein-ligand binding affinity. The model represents compounds based on both structural and bioactive properties and contains three feature extractors. Experimental results demonstrate the competitiveness and reliability of PLA-MoRe.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2022)
Article
Biochemical Research Methods
Kaili Wang, Renyi Zhou, Jing Tang, Min Li
Summary: To address the challenges in predicting protein-ligand binding affinity, we developed a novel graph neural network strategy called GraphscoreDTA, which combines graph neural network, bitransport information mechanism, and physics-based distance terms for the first time. Unlike other methods, GraphscoreDTA can effectively capture the mutual information between protein-ligand pairs and highlight the important atoms of the ligands and residues of the proteins. The results show that GraphscoreDTA outperforms existing methods on multiple test sets and demonstrates reliability in predicting protein-ligand binding affinity.
Review
Biochemistry & Molecular Biology
Lingling Zhao, Yan Zhu, Junjie Wang, Naifeng Wen, Chunyu Wang, Liang Cheng
Summary: This review provides a brief introduction to computation-based protein-ligand interactions (PLIs) and discusses various approaches, with a particular focus on machine learning methods. It also analyzes three research dynamics that can be further explored in future studies.
COMPUTATIONAL AND STRUCTURAL BIOTECHNOLOGY JOURNAL
(2022)
Article
Biochemical Research Methods
Sangmin Seo, Jonghwan Choi, Sanghyun Park, Jaegyoon Ahn
Summary: By proposing a deep-neural network model, we improved the prediction accuracy of protein-ligand complex binding affinity, with important features of descriptor embeddings and an attention mechanism. The proposed model outperformed existing models on most benchmark datasets.
BMC BIOINFORMATICS
(2021)
Review
Biochemical Research Methods
Ashwin Dhakal, Cole McKay, John J. Tanner, Jianlin Cheng
Summary: New drug production can take over 12 years and cost around $2.6 billion, with the COVID-19 pandemic emphasizing the need for more powerful computational methods in drug discovery. This review focuses on computational approaches using artificial intelligence (AI) to predict protein-ligand interactions, particularly in deep learning methods. The correlation between protein-ligand interaction aspects and the proposal to study them together could lead to more accurate machine learning-based prediction strategies.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Biochemical Research Methods
Kaili Wang, Renyi Zhou, Yaohang Li, Min Li
Summary: The study developed a deep learning method DeepDTAF for predicting protein-ligand binding affinity by integrating local and global contextual features. DeepDTAF showed significant accuracy improvement compared to state-of-art methods, making it a reliable tool for affinity prediction and drug discovery acceleration.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Biochemistry & Molecular Biology
Xun Wang, Dayan Liu, Jinfu Zhu, Alfonso Rodriguez-Paton, Tao Song
Summary: The study introduces a novel deep learning method CSConv2d for predicting protein-ligand interactions, showing better performance in binding affinity prediction compared to other models. Data experiments demonstrate the robustness of the proposed method in practice.
Article
Computer Science, Artificial Intelligence
Mei Li, Ye Cao, Xiaoguang Liu, Hua Ji
Summary: This article proposes a structure-aware graph attention diffusion network (SGADN) for efficient spatial structure learning of protein-ligand complexes by incorporating both distance and angle information. The SGADN utilizes line graph attention diffusion layers (LGADLs) to explore long-range bond node interactions and enhance the hierarchical structure learning, and also introduces an attentive pooling layer (APL) to refine the hierarchical structures in complexes.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Chemistry, Multidisciplinary
Bo Cai, Casey J. Krusemark
Summary: A novel assay method combining DNA encoding with split-and-pool sample handling is developed to improve small-molecule binding assays to target proteins. The approach involves affinity labeling of DNA-linked ligands to a protein target, allowing for quantification of DNA barcodes to detect ligand binding. This method demonstrates potential utility in high-throughput small-molecule screening.
ANGEWANDTE CHEMIE-INTERNATIONAL EDITION
(2022)
Article
Engineering, Chemical
Xinhao Che, Shiyang Chai, Zhongzhou Zhang, Lei Zhang
Summary: An improved blind docking method with a machine learning-based scoring function is proposed in this paper for the prediction of protein-ligand binding sites, and its excellent performance is demonstrated through two cases.
CHEMICAL ENGINEERING SCIENCE
(2022)
Article
Biology
Gaili Li, Yongna Yuan, Ruisheng Zhang
Summary: Accurately predicting protein-ligand binding affinities is crucial for understanding molecular properties. In this study, we propose the PLAsformer approach, which combines BiGRU, CNN, and attention mechanisms to effectively capture both local and global molecular information. The results demonstrate that PLAsformer outperforms current state-of-the-art methods for binding affinity prediction.
COMPUTATIONAL BIOLOGY AND CHEMISTRY
(2023)
Article
Biochemical Research Methods
Yu Wang, Zhengxiao Wei, Lei Xi
Summary: In this study, a new featurization method based on 3D convolutional neural network was proposed to generate a scoring function model. By testing four architectures and three featurization methods, and comparing with other scoring functions, the results showed that our model accurately and stably predicted the binding affinity of protein-ligand complexes. This model will contribute towards improving the success rate of virtual screening and accelerating the development of potential drugs or novel biologically active lead compounds.
BMC BIOINFORMATICS
(2022)
Article
Chemistry, Medicinal
Ai Shinobu, Chigusa Kobayashi, Yasuhiro Matsunaga, Yuji Sugita
Summary: Large-scale conformational transitions in multidomain proteins are crucial for their functions. The multi-basin G(o) over bar model provides an effective way to describe these transitions by exploring multiple potential pathways involving different intermediate structures. Application of this model to adenylate kinase demonstrates that optimized parameters allow the protein to transition frequently between various conformations.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2021)
Article
Biochemistry & Molecular Biology
Kiyoshi Yagi, Suyong Re, Takaharu Mori, Yuji Sugita
Summary: Recent advances in atomistic molecular dynamics simulations have allowed us to explore the conformational spaces of biomolecules and observe transitions between distinct structures. Weight average approaches are used to analyze various experimental measurements and improve our understanding of molecular functions based on atomic structures.
CURRENT OPINION IN STRUCTURAL BIOLOGY
(2022)
Article
Chemistry, Physical
Shingo Ito, Kiyoshi Yagi, Yuji Sugita
Summary: In this study, the allosteric regulation of TRPS was investigated using molecular dynamics simulations. The results revealed the effects of IGP binding on the conformation of the alpha and beta subunits, as well as the critical role of hydrogen bonds. These findings provide insights into the allosteric regulation of multidomain proteins.
JOURNAL OF PHYSICAL CHEMISTRY B
(2022)
Article
Biochemical Research Methods
Cheng Tan, Jaewoon Jung, Chigusa Kobayashi, Diego Ugarte La Torre, Shoji Takada, Yuji Sugita
Summary: Residue-level coarse-grained (CG) models play a crucial role in biomolecular simulations, enabling the study of large-scale biological phenomena with unified treatments of proteins and nucleic acids, as well as efficient parallel computations. The toolbox and methods introduced can help optimize the performance in CG MD simulations.
PLOS COMPUTATIONAL BIOLOGY
(2022)
Article
Biology
Hisham M. Dokainish, Suyong Re, Takaharu Mori, Chigusa Kobayashi, Jaewoon Jung, Yuji Sugita
Summary: The Spike protein plays a crucial role in neutralization and vaccine development for SARS-CoV-2. Its inherent flexibility provides essential information for designing antiviral drugs and vaccines.
Article
Chemistry, Medicinal
Hiraku Oshima, Yuji Sugita
Summary: The free-energy perturbation (FEP) method is an essential tool in in silico drug design, used to predict the free-energy changes of biomolecules in solvation and binding. However, conventional FEP requires computationally expensive reciprocal-space calculations. To address this limitation, this study proposes a modified Hamiltonian approach that introduces nonuniform scaling into the electrostatic potential, improving computational performance and avoiding the need for additional reciprocal-space calculations.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2022)
Article
Chemistry, Physical
Mao Oide, Yuji Sugita
Summary: In this study, a novel approach using the UMAP method was proposed to construct protein conformational landscapes, and native contact likelihood was used as feature variables to explore intermediate structures in protein folding. This method is useful for studying large-scale conformational changes in biomacromolecules.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Biophysics
Hisham M. Dokainish, Yuji Sugita
Summary: A single mutation (D614G) in the spike protein of SARS-CoV-2, which causes COVID-19, has become the dominant variant. This mutation enhances the virus's infectivity by inducing structural changes. Molecular dynamics simulations reveal that the mutation orders the structure, weakens local interactions, and alters global interactions, leading to conformational changes. Understanding this mutation is crucial as it is present in all variants of concern.
BIOPHYSICAL JOURNAL
(2023)
Review
Biophysics
Yasuhiro Matsunaga, Motoshi Kamiya, Hiraku Oshima, Jaewoon Jung, Shingo Ito, Yuji Sugita
Summary: MBAR is a widely used method in analyzing MD simulation data to estimate free-energy changes between different states and averaged properties. Due to its broad applicability, the MBAR equations are rather difficult to apply for free-energy calculations using different types of MD simulations.
BIOPHYSICAL REVIEWS
(2022)
Article
Chemistry, Multidisciplinary
Jaewoon Jung, Chigusa Kobayashi, Yuji Sugita
Summary: gREST is an enhanced sampling algorithm used for proteins or other systems with rugged energy landscapes. It differs from the REMD method as the solvent temperatures are the same in all replicas, while solute temperatures are different and frequently exchanged between replicas to explore various solute structures. The gREST scheme is applied to large biological systems using a large number of processors, reducing communication time and performing energy evaluations on-the-fly during simulations. These advanced schemes in the GENESIS software provide new possibilities for studying large biomolecular complex systems with slow conformational dynamics.
JOURNAL OF COMPUTATIONAL CHEMISTRY
(2023)
Article
Chemistry, Multidisciplinary
Cheng Tan, Ai Niitsu, Yuji Sugita
Summary: In this study, the interactions between Hero11 protein and TDP-43-LCD protein were investigated using multiscale molecular dynamics simulations. Three possible regulatory mechanisms of Hero11 were proposed based on the simulation results. It was found that Hero11 can permeate into TDP-43-LCD condensates and induce changes in their conformation, intermolecular interactions, and dynamics. These mechanisms provide new insights into the regulation of biomolecular condensation under different conditions.
Article
Chemistry, Physical
Diego Ugarte La Torre, Shoji Takada, Yuji Sugita
Summary: iSoLF is a coarse-grained model used in molecular dynamics simulations of biological membranes with the implicit solvent approximation. By explicitly treating the electrostatic interactions and adding new particle types, the model can now simulate a wider range of lipid molecules. The improved model has been parameterized and validated, and it is also capable of simulating phase behaviors of lipid mixtures.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Multidisciplinary
Isseki Yu, Takaharu Mori, Daisuke Matsuoka, Donatas Surblys, Yuji Sugita
Summary: The rapid increase in computational power has enabled atomistic molecular dynamics simulations of biomolecules in large-scale biological systems. Spatial decomposition analysis (SPANA) reduces computational time and memory size by distributing analysis tasks to multiple CPU cores, opening new possibilities for detailed atomistic analyses of biomacromolecules and other molecules in MD simulation trajectories.
JOURNAL OF COMPUTATIONAL CHEMISTRY
(2023)
Article
Biochemistry & Molecular Biology
Daiki Matsubara, Kento Kasahara, Hisham M. Dokainish, Hiraku Oshima, Yuji Sugita
Summary: Proper balance between protein-protein and protein-water interactions is crucial for molecular dynamics simulations of proteins. Increasing the protein-water interactions helps optimize the balance and has significant effects on diffusive properties of proteins in crowded solutions.