Article
Chemistry, Physical
Srujan Sapkal, Balasubramanian Kandasubramanian, Prashant Dixit, Himanshu Sekhar Panda
Summary: In this study, a machine learning model is developed to predict the functional properties of KNN-based ceramics. By identifying important features, the design process can be accelerated and synthesis methods can be optimized. The experimental results are consistent with the predictions, indicating that the model has good predictive performance and applicability.
MATERIALS TODAY ENERGY
(2023)
Article
Biology
Lasse M. Blaabjerg, Maher M. Kassem, Lydia L. Good, Nicolas Jonsson, Matteo Cagiada, Kristoffer E. Johansson, Wouter Boomsma, Amelie Stein, Kresten Lindorff-Larsen
Summary: This study presents a method called RaSP, which utilizes deep learning representations to predict protein stability changes accurately and rapidly. The method performs well in saturation mutagenesis stability predictions and has been applied to large-scale stability analysis of the human proteome, revealing important insights into the role of protein stability in genetic diseases.
Article
Physics, Multidisciplinary
Zhongyu Wan, Quan-De Wang, Dongchang Liu, Jinhu Liang
Summary: This study demonstrates the use of machine learning models and descriptors to efficiently predict the properties of new quantum materials, showing excellent performance in predicting the electronic structure properties of topological insulators.
Article
Chemistry, Multidisciplinary
Andrew Ma, Yang Zhang, Thomas Christensen, Hoi Chu Po, Li Jing, Liang Fu, Marin Soljacic
Summary: By utilizing machine learning, we have developed a simple chemical rule that accurately diagnoses whether a material is topological or not, based solely on its chemical formula. We have also established a high-throughput procedure for discovering topological materials using this heuristic rule, followed by ab initio validation. This approach has led to the discovery of new topological materials that cannot be identified using symmetry indicators, some of which show promise for experimental observation.
Article
Materials Science, Multidisciplinary
H. M. Yuan, S. H. Han, R. Hu, W. Y. Jiao, M. K. Li, H. J. Liu, Y. Fang
Summary: This paper proposes a neural network model that can quickly evaluate the Seebeck coefficient at arbitrary carrier concentration. By using only a few elemental properties as input features, the model exhibits high correlation between the real and predicted Seebeck coefficients. The model can be used to screen Heusler compounds for accelerated discovery of new materials with desired Seebeck coefficients.
MATERIALS TODAY PHYSICS
(2022)
Article
Computer Science, Artificial Intelligence
Mykola Galushka, Chris Swain, Fiona Browne, Maurice D. Mulvenna, Raymond Bond, Darren Gray
Summary: This study presents a new deep learning model for conducting preliminary screening of chemical compounds in-silico and accurately predicting their properties. This approach has the potential to provide a more efficient pathway for pharmaceutical companies to discover new medications.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Materials Science, Multidisciplinary
Jonggul Lee, Jungho Shin, Tae-Wook Ko, Seunghee Lee, Hyunju Chang, YunKyong Hyon
Summary: This study systematically compared two state-of-the-art frameworks, CGCNNs and SISSO, for predicting properties of crystalline materials. Both models have the advantage of not requiring painstakingly handcrafted descriptors, but differ in their use of structure information and interpretability.
MATERIALS RESEARCH EXPRESS
(2021)
Article
Chemistry, Multidisciplinary
Shidang Xu, Xiaoli Liu, Pengfei Cai, Jiali Li, Xiaonan Wang, Bin Liu
Summary: This study establishes a database of AIE/ACQ molecules and predicts the properties of molecules in the aggregated state using machine learning models, proposing a multi-modal approach and developing an ensemble strategy to achieve reasonable prediction results in an unknown molecular space.
Article
Biochemistry & Molecular Biology
Firda Nurul Auliah, Andi Nur Nilamyani, Watshara Shoombuatong, Md Ashad Alam, Md Mehedi Hasan, Hiroyuki Kurata
Summary: The PUP-Fuse is a new prediction model for pupylation site prediction that integrates multiple sequence representations and achieves good prediction results based on machine learning algorithms.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2021)
Article
Food Science & Technology
Yu Song, Sihao Chang, Jing Tian, Weihua Pan, Lu Feng, Hongchao Ji
Summary: This study explores taste prediction using various molecular feature representations and evaluates the performance of different machine learning algorithms. The results show that GNN-based models outperform other approaches, with the molecular fingerprints + GNN consensus model performing the best.
Article
Chemistry, Physical
Kyungtaek Lee, Young In Jhon, Suh-young Kwon, Geunweon Lim, Jeehwan Kim, Ju Han Lee
Summary: This study investigates the nonlinear refraction and absorption properties of SnTe TCIs for the first time, revealing their high nonlinear refractive index and nonlinear absorption coefficient. Ab-initio simulations demonstrate the suitability of SnTe TCIs for ultrabroad-band applications and mechanical stability at high thermal conditions. Furthermore, the study shows that 1550 nm mode-locked femtosecond lasers can be achieved using the nonlinear optical absorption of SnTe TCIs without nano-engineered structures like quantum dots.
JOURNAL OF ALLOYS AND COMPOUNDS
(2022)
Article
Physics, Condensed Matter
Taruto Atsumi, Kosei Sato, Yudai Yamaguchi, Masato Hamaie, Risa Yasuda, Naoto Tanibata, Hayami Takeda, Masanobu Nakayama, Masayuki Karasuyama, Ichiro Takeuchi
Summary: Materials informatics plays a crucial role in the efficient discovery and development of new functional materials. Machine-learning regression techniques are commonly used to establish the correlation between material properties. The prediction performance can be improved by using compositional descriptors.
PHYSICA STATUS SOLIDI B-BASIC SOLID STATE PHYSICS
(2022)
Review
Chemistry, Multidisciplinary
Ziyang Fu, Weiyi Liu, Chen Huang, Tao Mei
Summary: Rational design and discovery of materials have a huge impact on technology and society, requiring a multi-stage design process including material composition, structure, properties, and process design. Machine learning is considered a new way to explore the materials field by finding correlations in material properties across data.
Article
Materials Science, Multidisciplinary
Jacob Peloquin, Alina Kirillova, Cynthia Rudin, L. C. Brinson, Ken Gall
Summary: This research proposes a framework for quickly predicting key mechanical properties of 3D printed gyroid lattices using information about the base material and porosity of the structure. The performance of the model was then compared to numerical simulation data and demonstrated similar accuracy at a fraction of the computation time. The model development serves as an advancement in ML-driven mechanical property prediction that can be used to guide extension of current and future models.
MATERIALS & DESIGN
(2023)
Article
Chemistry, Physical
Logan Williams, Arpan Mukherjee, Ruhil Dongol, Krishna Rajan
Summary: This paper presents a lightweight neural network that uses Hirshfeld surfaces as material representation to predict the properties of hybrid organic-inorganic perovskites (HOIPs). The network utilizes a few Hirshfeld surface features along with qualitative and quantitative variables to achieve fast prediction of HOIP properties. The effectiveness of this approach is demonstrated through a comparison with other methods using different compound chemistries.
JOURNAL OF PHYSICAL CHEMISTRY C
(2023)
Article
Chemistry, Multidisciplinary
Jasiel O. Strubbe-Rivera, Jiahui Chen, Benjamin A. West, Kristin N. Parent, Guo-Wei Wei, Jason N. Bazil
Summary: Mitochondrial cristae are dynamic invaginations of the inner membrane that play a key role in ATP production. Structural alterations caused by genetic abnormalities or detrimental environmental factors can reduce mitochondrial metabolic capacity. A computational strategy was proposed to understand how cristae are formed and how calcium phosphate granules affect mitochondrial energy metabolism.
APPLIED SCIENCES-BASEL
(2021)
Article
Biology
Rui Wang, Jiahui Chen, Kaifu Gao, Yuta Hozumi, Changchuan Yin, Guo-Wei Wei
Summary: The study reveals the presence of four sub-strains and eleven top mutations in the United States, with five and eight concurrent mutations prevailing in two groups, while another group with three concurrent mutations gradually fading out. Additionally, it is found that female immune systems are more active than those of males in responding to SARS-CoV-2 infections.
COMMUNICATIONS BIOLOGY
(2021)
Article
Biochemistry & Molecular Biology
Jiahui Chen, Kaifu Gao, Rui Wang, Guo-Wei Wei
Summary: The ongoing vaccination and development of intervention offer hope to end the global COVID-19 pandemic, but emerging SARS-CoV-2 variants could compromise existing vaccines and antibody therapies. Studies on potential threats from mutations are limited, and the impact on clinical trial antibodies is largely unknown.
JOURNAL OF MOLECULAR BIOLOGY
(2021)
Article
Multidisciplinary Sciences
Dong Chen, Kaifu Gao, Duc Duy Nguyen, Xin Chen, Yi Jiang, Guo-Wei Wei, Feng Pan
Summary: Researchers proposed an algebraic graph-assisted bidirectional transformer framework, which can integrate massive unlabeled molecular data into molecular representations via a self-supervised learning strategy and incorporate 3D stereochemical information from graphs, showing state-of-the-art performance in molecular property prediction.
NATURE COMMUNICATIONS
(2021)
Article
Biochemistry & Molecular Biology
Qing Zhu, Yuzhe Du, Yoshiko Nomura, Rong Gao, Zixuan Cang, Guo-Wei Wei, Dalia Gordon, Michael Gurevitz, James Groome, Ke Dong
Summary: The study indicates that charge substitutions in different structural domains of the sodium channel can enhance the activity of scorpion toxins, particularly the charge reversal substitutions in the voltage-sensing modules of domain III which can facilitate the actions of toxins on IIS4 or IVS4 voltage sensors.
INSECT BIOCHEMISTRY AND MOLECULAR BIOLOGY
(2021)
Article
Biology
Jiahui Chen, Yuchi Qiu, Rui Wang, Guo-Wei Wei
Summary: Due to its high transmissibility, Omicron BA.1 became the dominant variant in late 2021, replacing the Delta variant, and was later replaced by the even more transmissible Omicron BA.2. This study tackles the challenge of capturing both topological change and homotopic shape evolution in virus-human protein-protein binding using persistent Laplacian-based deep learning models. The analysis reveals that BA.4 and BA.5 are more infectious than BA.2 and are projected to become new dominant variants. Additionally, the proposed models outperform state-of-the-art methods in predicting mutation-induced protein-protein binding free energy changes.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Biology
Hongsong Feng, Guo-Wei Wei
Summary: In this study, machine learning-based in silico tools were used to screen compounds in the DrugBank database. It was found that 227 out of 8641 DrugBank compounds potentially block the hERG channel, which may lead to serious drug safety issues.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Biology
Hongsong Feng, Jian Jiang, Guo-Wei Wei
Summary: Opioid use disorder (OUD) is a chronic and relapsing condition characterized by continued and compulsive use of opioids despite harmful consequences. Drug repurposing using machine learning is an efficient and cost-effective approach for discovering medications for OUD treatment.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Biology
Hongsong Feng, Rana Elladki, Jian Jiang, Guo-Wei Wei
Summary: Opioid use disorder (OUD) is a global public health issue, and the efficacy of current treatment options needs to be improved. This study utilized machine learning and protein-protein interaction networks to explore potential drug candidates for OUD treatment. The findings provide valuable insights and promising tools for the development of pharmacological treatments.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Chemistry, Multidisciplinary
Xiaoqi Wei, Jiahui Chen, Guo-Wei Wei
Summary: Persistent topological Laplacians (PTLs) are a new tool in topological data analysis for studying protein structural changes. By using PTLs, we can reveal the spectrum changes in protein structures among SARS-CoV-2 variants and analyze the structural changes induced by RBD and ACE2 binding. Furthermore, PTLs can be utilized in a topological deep learning paradigm and for predictions of deep mutational scanning datasets for SARS-CoV-2 variants.
JOURNAL OF COMPUTATIONAL BIOPHYSICS AND CHEMISTRY
(2023)
Review
Biochemical Research Methods
Yuchi Qiu, Guo-Wei Wei
Summary: Protein engineering is a promising field in biotechnology with the potential to revolutionize various areas. Machine learning models, particularly those based on natural language processing, have greatly accelerated protein engineering by leveraging protein databases. Advances in topological data analysis and artificial intelligence-based protein structure prediction have enabled more powerful structure-based machine learning-assisted protein engineering strategies. This review provides a comprehensive and indispensable set of methodological components, including topological data analysis and natural language processing, to facilitate the future development of protein engineering.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biology
Jiahui Chen, Daniel R. Woldring, Faqing Huang, Xuefei Huang, Guo-Wei Wei
Summary: High-throughput deep mutational scanning (DMS) experiments have revolutionized various fields such as protein engineering, drug discovery, immunology, cancer biology, and evolutionary biology by providing systematic understanding of protein functions. However, the enormous mutational space associated with proteins exceeds current experimental capabilities, necessitating alternative approaches for DMS. In this study, we propose a topological deep learning (TDL) paradigm that utilizes a new topological data analysis (TDA) technique based on the persistent spectral theory. Our results demonstrate the accuracy and reliability of the TDL-DMS model in predicting binding interface mutations using SARS-CoV-2 datasets. This finding has significant implications for SARS-CoV-2 variant forecasting, antibody design, vaccine development, precision medicine, and protein engineering.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Chemistry, Multidisciplinary
Jiahui Chen, Kaifu Gao, Rui Wang, Guo-Wei Wei
Summary: This study examines the impact of mutations on the spike protein of COVID-19, particularly on vaccines and antibody therapies. The research findings reveal that certain mutations may weaken the binding between the spike protein and antibodies, potentially reducing the efficacy of current treatments. Moreover, it is discovered that some mutations could enhance the binding between the spike protein and human angiotensin-converting enzyme 2 (ACE2), leading to more infectious variants of the virus.
Review
Chemistry, Physical
Duc Duy Nguyen, Zixuan Cang, Guo-Wei Wei
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2020)