Article
Green & Sustainable Science & Technology
Pratima Kumari, Durga Toshniwal
Summary: A new ensemble model, XGBF-DNN, is proposed for hourly global horizontal irradiance forecast, integrating extreme gradient boosting forest and deep neural networks with ridge regression. The model shows stability and high prediction accuracy across various climatic conditions, making it suitable for solar energy system design and planning.
JOURNAL OF CLEANER PRODUCTION
(2021)
Article
Materials Science, Multidisciplinary
Seungro Lee, Joonhee Park, Naksoo Kim, Taeyong Lee, Luca Quagliato
Summary: This paper presents a machine learning methodology that can learn from simulation results, experimental data, or sensor signals, and is capable of predicting and optimizing specific user-defined process and design parameters. The methodology utilizes an enhanced Extreme Gradient Boosting (XGB) algorithm and a metaheuristic search algorithm based on Differential Evolution (DE) architecture for optimization.
MATERIALS & DESIGN
(2023)
Article
Construction & Building Technology
Hieu Nguyen, Nhat-Duc Hoang
Summary: This paper presents alternative solutions for classifying concrete spall severity based on computer vision approaches. XGBoost optimized by the Aquila metaheuristic and used with ARCS-LBP achieved an outstanding classification performance with a classification accuracy rate of roughly 99% for real-world concrete surface images.
AUTOMATION IN CONSTRUCTION
(2022)
Article
Biochemistry & Molecular Biology
Omar Alghushairy, Farman Ali, Wajdi Alghamdi, Majdi Khalid, Raed Alsini, Othman Asiry
Summary: The identification of druggable proteins is crucial for drug development, personalized medicine, and understanding disease mechanisms. This study introduces a computational predictor called Drug-LXGB, which utilizes machine learning strategies to enhance the identification of druggable proteins. The predictor achieved high predictive accuracy through feature selection algorithms and learning methods.
JOURNAL OF BIOMOLECULAR STRUCTURE & DYNAMICS
(2023)
Article
Computer Science, Information Systems
Abdoulie Fatty, An-Jui Li, Zhi-Guang Qian
Summary: In this study, an artificial intelligence-based technique for rock slope stability prediction is proposed. The model achieves impressive performance on a comprehensive rock slope database and the SHAP algorithm is used to interpret the dependencies between input and output variables.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Engineering, Geological
Aali Pant, G. V. Ramana
Summary: This study proposes the prediction of pullout interaction coefficient of geogrids using data-driven machine learning regression algorithms, primarily focusing on extreme gradient boosting (XGBoost) method. The XGBoost model shows significantly superior and robust prediction compared to the random forest (RF) model, with an accuracy of 85% and 77% respectively. The importance analysis identifies normal stress as the most significant factor influencing the pullout interaction coefficients.
GEOTEXTILES AND GEOMEMBRANES
(2022)
Article
Green & Sustainable Science & Technology
Odey Alshboul, Ali Shehadeh, Ghassan Almasabha, Ali Saeed Almuflih
Summary: Accurate prediction of green building costs is crucial for decision-making and management. This study presents machine learning-based algorithms for cost prediction and evaluates their accuracy.
Article
Ophthalmology
Yunfei Li, Jingyu Ma, Jun Xiao, Yujiao Wang, Weimin He
Summary: The purpose of this study was to develop a machine learning model to evaluate the activity stage of extraocular muscles in thyroid-associated ophthalmopathy (TAO). Three machine learning models were constructed and compared for their performance in diagnosing TAO. The LightGBM model showed the best diagnostic performance, and several features were found to have a significant impact on the model's prediction.
GRAEFES ARCHIVE FOR CLINICAL AND EXPERIMENTAL OPHTHALMOLOGY
(2023)
Article
Engineering, Multidisciplinary
Anurag Malik, Mandeep Kaur Saggi, Sufia Rehman, Haroon Sajjad, Samed Inyurt, Amandeep Singh Bhatia, Aitazaz Ahsan Farooque, Atheer Y. Oudah, Zaher Mundher Yaseen
Summary: In this study, innovative techniques of Deep Learning and Gradient Boosting Machine models were developed for modeling monthly pan evaporation based on maximum air temperature. The results showed that the Deep Learning model outperformed the Gradient Boosting Machine model in predicting the monthly pan evaporation for both the Kiashahr and Ranichauri stations. The study highlights the superior performance of Deep Learning-based models in environmental modeling.
ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS
(2022)
Article
Engineering, Biomedical
J. Mateo, J. M. Rius-Peris, A. I. Marana-Perez, A. Valiente-Armero, A. M. Torres
Summary: The study proposed a medical treatment identification prediction model based on XGB machine learning method, with XGB showing the best prediction accuracy among the compared methods, while others also demonstrated moderate prediction accuracy and sensitivity.
BIOCYBERNETICS AND BIOMEDICAL ENGINEERING
(2021)
Article
Construction & Building Technology
Ji-Gang Xu, Shi-Zhi Chen, Wei-Jie Xu, Zi-Sen Shen
Summary: Accurate prediction of the shear strength of old and new concrete interfaces is crucial for the design and assessment of precast and retrofitted concrete structures. This study developed an explainable machine learning model for predicting interface shear strength, with the XGBoost model showing the best performance among the ML-models and outperforming empirical models. Key factors affecting predictions include reinforcement ratio, surface type, interface section width, and concrete strength.
CONSTRUCTION AND BUILDING MATERIALS
(2021)
Article
Computer Science, Artificial Intelligence
Jiaming Han, Kunxin Shu, Zhenyu Wang
Summary: Annual increases in global energy consumption are inevitable due to a growing global economy and population. Among sectors, the construction industry consumes 20.1% of the world's total energy, making it crucial to explore methods for estimating energy usage. Various computational approaches exist, including statistics-based, engineering-based, and machine learning-based methods. Machine learning-based frameworks outperform the others. In our study, we propose using the Extreme Gradient Boosting algorithm to predict energy consumption, achieving better results with combined historical and date features.
PEERJ COMPUTER SCIENCE
(2023)
Article
Computer Science, Artificial Intelligence
Jessica Tito Vieira, Robson Bruno Dutra Pereira, Carlos Henrique Lauro, Lincoln Cardoso Brandao, Joao Roberto Ferreira
Summary: This study presents a statistical learning approach for modeling and optimization of the internal turning process in PEEK tubes, and an experimental multi-objective evolutionary optimization method. The results indicate that the extreme gradient boosting model has advantages in prediction and optimization.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Medicine, Research & Experimental
Thomas R. Lane, Fabio Urbina, Laura Rank, Jacob Gerlach, Olga Riabova, Alexander Lepioshkin, Elena Kazakova, Anthony Vocat, Valery Tkachenko, Stewart Cole, Vadim Makarov, Sean Ekins
Summary: Overall, tuberculosis remains a major global health challenge requiring the development of new drug treatments. Machine learning approaches have been utilized to identify new active compounds and develop classification models. New models for scoring compound libraries and visualizing molecules in chemical property space have been provided.
MOLECULAR PHARMACEUTICS
(2022)
Article
Computer Science, Theory & Methods
Jemin Lee, Misun Yu, Yongin Kwon, Taeho Kim
Summary: To adopt convolutional neural networks (CNN) for resource-constrained targets, compressing the CNN models through quantization is necessary. Previous research proposed post-training quantization methods to address issues such as dataset sensitivity, high computational requirements, and time consumption. Additional methods like calibration, clipping, and mixed-precision were proposed to compensate for accuracy drop without retraining. However, exhaustive or heuristic searches are impractical. To solve this, the authors propose an auto-tuner called Quantune, which uses gradient tree boosting to accelerate the search for quantization configurations and reduce quantization error.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2022)
Article
Biochemistry & Molecular Biology
Gaoqi Weng, Xuanyan Cai, Dongsheng Cao, Hongyan Du, Chao Shen, Yafeng Deng, Qiaojun He, Bo Yang, Dan Li, Tingjun Hou
Summary: PROTAC-DB 2.0 is an updated online database that contains structural and experimental data about PROTACs. This second version expands the number of PROTACs to 3270 and provides additional information to aid in the understanding and design of PROTACs.
NUCLEIC ACIDS RESEARCH
(2023)
Review
Chemistry, Multidisciplinary
Gaoang Wang, Lei Xu, Haiyi Chen, Yifei Liu, Peichen Pan, Tingjun Hou
Summary: This article summarizes recent and representative studies on voltage-gated sodium channels (VGSCs/Na(v)s) from the perspective of computer-aided drug design (CADD) and molecular modeling. It covers the structural biology of VGSCs, virtual screening and drug design based on CADD, and functional studies using molecular modeling technologies. The article concludes the achievements in the field of VGSCs and discusses the shortcomings found in previous studies.
WILEY INTERDISCIPLINARY REVIEWS-COMPUTATIONAL MOLECULAR SCIENCE
(2023)
Article
Biochemical Research Methods
Shukai Gu, Chao Shen, Jiahui Yu, Hong Zhao, Huanxiang Liu, Liwei Liu, Rong Sheng, Lei Xu, Zhe Wang, Tingjun Hou, Yu Kang
Summary: This study evaluated the impact of structural dynamic information on binding affinity prediction and found that the optimized molecular dynamics protocol improved the predictive performance for the TAF1-BD2 target with high structural flexibility, but not for the less flexible JAK1 and DDR1 targets.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Chemistry, Multidisciplinary
Lei Xu, Xuyang Xuyan, Minmin Minmi, Zhiji Wang, Chao Shee, Qianwen Mu, Bo Feng, Yechun Xu, Tingjun Hou, Lixin Gao, Haini Jiang, Jia Li, Yubo Zhou, Wenlong Wang
Summary: In this study, a new class of thieno[2,3-b]quinolineprocaine hybrid molecules were reported as allosteric activators of SHP-1. The representative hybrid compound 3b displayed SHP-1 activating effect with an EC50 of 5.48 ± 0.28 μmol/L. Further investigations confirmed that 3b allosterically interacted with SHP-1, switched it from close to open conformation, blocked SHP-1/p STAT3 pathway, induced apoptosis and inhibited ABC-DLBCL cell proliferation.
CHINESE CHEMICAL LETTERS
(2023)
Article
Chemistry, Medicinal
Teng-Zhi Long, Shao-Hua Shi, Shao Liu, Ai-Ping Lu, Zhao-Qian Liu, Min Li, Ting-Jun Hou, Dong-Sheng Cao
Summary: This study constructed a high-quality dataset and established a series of classification models using machine learning algorithms to predict hematotoxicity. The best model based on Attentive FP showed excellent performance on both the validation and test sets. Additionally, the study utilized SHAP and atom heatmap methods to identify important features and structural fragments related to hematotoxicity, and employed MMPA and representative substructure derivation technique to further investigate the transformation principles and distinctive structural features of hematotoxic chemicals.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Chemistry, Physical
Yihao Zhao, Jintu Zhang, Haotian Zhang, Shukai Gu, Yafeng Deng, Yaoquan Tu, Tingjun Hou, Yu Kang
Summary: Inspired by GaMD, this work proposes a new accelerated molecular dynamics method called Sigmoid accelerated molecular dynamics (SaMD), which improves the balance between the highest acceleration and accurate reweighting by adding a Sigmoid boost potential. Compared with GaMD, SaMD extends the accessible time scale and improves computational efficiency, and it achieves better results in alanine dipeptide, chignolin folding, and protein-ligand binding tasks.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2023)
Article
Pharmacology & Pharmacy
Jialu Wu, Yue Wan, Zhenxing Wu, Shengyu Zhang, Dongsheng Cao, Chang-Yu Hsieh, Tingjun Hou
Summary: MF-SuP-pKa is a novel pKa prediction model that utilizes subgraph pooling, multi-fidelity learning, and data augmentation. The model captures the local and global environments around ionization sites for micro-pKa prediction using a knowledge-aware subgraph pooling strategy. By fitting low-fidelity data to high-fidelity data through transfer learning, MF-SuP-pKa achieves superior performance compared to state-of-the-art models with less high-fidelity training data.
ACTA PHARMACEUTICA SINICA B
(2023)
Article
Chemistry, Multidisciplinary
Rongfan Tang, Zhe Wang, Sutong Xiang, Lingling Wang, Yang Yu, Qinghua Wang, Qirui Deng, Tingjun Hou, Huiyong Sun
Summary: Proteolysis-targeting chimeras (PROTACs) selectively degrade target proteins and are an attractive technology in drug discovery. In this study, the kinetic mechanism of PROTAC MZ1 targeting the bromodomain (BD) of BET protein and von Hippel-Lindau E3 ligase (VHL) was characterized and analyzed using simulations and free energy calculations. The results showed that MZ1 prefers to bind with E3 ligase in the formation of the target-PROTAC-E3 ligase ternary complex. The binding characteristics revealed in this study may accelerate the rational design of PROTACs with higher degradation efficiency.
Article
Biochemical Research Methods
Xujun Zhang, Chao Shen, Tianyue Wang, Yafeng Deng, Yu Kang, Dan Li, Tingjun Hou, Peichen Pan
Summary: Cracking the code of protein-ligand interaction is crucial for drug design and discovery. The ML-based PLI capturer (ML-PLIC) is a web platform that automatically characterizes PLI and generates machine learning-based scoring functions to identify potential binders. It outperforms traditional docking tools and performs competitively with deep learning-based methods. ML-PLIC integrates physical and biological knowledge to design a structure-based virtual screening pipeline.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Jian Gao, Zheyuan Shen, Yufeng Xie, Jialiang Lu, Yang Lu, Sikang Chen, Qingyu Bian, Yue Guo, Liteng Shen, Jian Wu, Binbin Zhou, Tingjun Hou, Qiaojun He, Jinxin Che, Xiaowu Dong
Summary: This study introduces a more elegant transformer-based framework, TransFoxMol, to improve the artificial intelligence's understanding of molecular structure-property relationships. Experimental results show that TransFoxMol achieves state-of-the-art performance and outperforms baseline models on small-scale datasets.
BRIEFINGS IN BIOINFORMATICS
(2023)
Review
Pharmacology & Pharmacy
Shukai Gu, Huanxiang Liu, Liwei Liu, Tingjun Hou, Yu Kang
Summary: Kinases play a crucial role in cellular processes and accurate kinase-profiling prediction is vital for drug discovery. This review provides an overview of the latest advancements in machine learning and deep learning models for kinase profiling, discussing the challenges and future directions in this field.
DRUG DISCOVERY TODAY
(2023)
Article
Chemistry, Medicinal
Zhe Wang, Haiyang Zhong, Jintu Zhang, Peichen Pan, Dong Wang, Huanxiang Liu, Xiaojun Yao, Tingjun Hou, Yu Kang
Summary: This study systematically evaluates the performance of traditional methods and AI models in small-molecule conformer generation. The results show that traditional methods outperform AI models in reproducing bioactive conformations, while an AI model has an advantage in generating low-energy conformations.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Chemistry, Multidisciplinary
Xujun Zhang, Chao Shen, Dejun Jiang, Jintu Zhang, Qing Ye, Lei Xu, Tingjun Hou, Peichen Pan, Yu Kang
Summary: Machine learning-based scoring functions (MLSFs) have the potential to improve virtual screening capabilities compared to classical scoring functions (SFs). However, the high computational cost and limited descriptors used in MLSFs and protein-ligand interaction characterization may impact accuracy and efficiency. In this study, a new SF called TB-IECS was proposed, combining energy terms from Smina and NNScore version 2 using the XGBoost algorithm. TB-IECS outperformed classical SFs and balanced efficiency and accuracy for practical virtual screening.
JOURNAL OF CHEMINFORMATICS
(2023)
Article
Chemistry, Multidisciplinary
Zipeng Zhong, Jie Song, Zunlei Feng, Tiantao Liu, Lingxiang Jia, Shaolun Yao, Tingjun Hou, Mingli Song
Summary: Retrosynthesis is the cornerstone of organic chemistry, and recent advances in deep learning and artificial intelligence have revolutionized the field. This comprehensive review provides a taxonomy and evaluation of existing methods, as well as an introduction to popular databases and platforms for retrosynthesis.
WILEY INTERDISCIPLINARY REVIEWS-COMPUTATIONAL MOLECULAR SCIENCE
(2023)
Article
Computer Science, Artificial Intelligence
Odin Zhang, Jintu Zhang, Jieyu Jin, Xujun Zhang, Renling Hu, Chao Shen, Hanqun Cao, Hongyan Du, Yu Kang, Yafeng Deng, Furui Liu, Guangyong Chen, Chang-Yu Hsieh, Tingjun Hou
Summary: This article introduces a three-dimensional molecular generative model called ResGen, which is conditioned on protein pockets and can design organic molecules inside a given target. The ResGen model has a higher success rate in generating novel molecules that bind more tightly to unseen targets than existing approaches. It also successfully generates drug-like molecules with lower binding energy and higher diversity than state-of-the-art methods in real-world scenarios.
NATURE MACHINE INTELLIGENCE
(2023)