Article
Computer Science, Artificial Intelligence
Hengtong Zhang, Yaliang Li, Bolin Ding, Jing Gao
Summary: Due to the openness of online platforms, recommendation systems are vulnerable to data poisoning attacks that manipulate their results. Existing attack methods are either based on heuristic rules with unsatisfactory performance, or they require strong knowledge of the target system. In this paper, we propose LOKI, a practical poisoning attack approach, which utilizes reinforcement learning to generate user behavior samples for data poisoning. By interacting with a recommender simulator, LOKI leverages the transferability of adversarial samples to poison the target system. Extensive experiments show that LOKI outperforms existing methods and we discuss the characteristics of vulnerable users/items and the use of anomaly detection methods to mitigate the impact of data poisoning attacks.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Mathematics, Applied
Julius Damarackas, Vygantas Paulauskas
Summary: This paper explores the summation of values of a real-valued stationary random field over rectangles with a focus on stable and unstable behavior. The introduction of Lamperti type theorem helps classify these behaviors, and examples on Z(2) and Z(3) are provided to demonstrate the scaling transition phenomenon associated with unstable behavior.
JOURNAL OF MATHEMATICAL ANALYSIS AND APPLICATIONS
(2021)
Article
Engineering, Civil
Erhao Meng, Shengzhi Huang, Qiang Huang, Wei Fang, Hao Wang, Guoyong Leng, Lu Wang, Hao Liang
Summary: This study introduces a stepwise decomposition sampling (SDS) strategy and an innovative input selection framework for building a strong decomposition-based monthly streamflow prediction model. Results show that combining these strategies can improve prediction accuracy and reliability, making it a useful and powerful tool for practical hydrological prediction work in the context of climate change.
WATER RESOURCES MANAGEMENT
(2021)
Article
Pharmacology & Pharmacy
Chi So, Lap Y. Leung, Ariel R. Muliadi, Ajit S. Narang, Chen Mao
Summary: The study established a virtual tool for predicting pharmaceutical roller compaction, demonstrating a critical nip angle and simplifying the model without considering friction angle, with good predictive performance.
INTERNATIONAL JOURNAL OF PHARMACEUTICS
(2021)
Article
Chemistry, Physical
YingXing Cheng, Toon Verstraelen
Summary: This study proposes an extended version of the frequency-dependent polarizable force field ACKS2, called ACKS2 omega, which allows for theoretical predictions of the dynamical response properties of finite systems. Parameters are computed as expectation values of an electronic wavefunction, and the hardness matrix is reused from ACKS2 using an adiabatic approximation. Numerical validation shows that accurate models can be obtained using atomic monopoles and dipoles.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Statistics & Probability
Mitia Duerinckx, Julian Fischer, Antoine Gloria
Summary: This paper studies the fluctuations of the solution-operator of a linear elliptic partial differential equation with a random coefficient field, and establishes the scaling limit and (non)degeneracy of the fluctuations, extending previous results.
ANNALS OF APPLIED PROBABILITY
(2022)
Article
Engineering, Electrical & Electronic
Firdous A. Shah, Aajaz A. Teali
Summary: In this article, a novel scaling Wigner distribution is introduced, which combines the advantages of fractional instantaneous auto-correlation and the linear canonical transform. The study shows that this distribution performs exceptionally well in detecting linear frequency modulated signals, while maintaining a good time-frequency resolution and clear auto-terms angle resolution.
CIRCUITS SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Multidisciplinary
Ilias N. Giannakeas, Zahra Sharif Khodaei, M. H. Aliabadi
Summary: This paper presents a novel framework for compensating the effect of temperature on guided wave structural health monitoring systems. The proposed methodology updates compensation factors using observations obtained at lower scales and propagates the estimated factors to higher scales within a Bayesian framework. The results show that the proposed methodology improves the fidelity of the compensation algorithm and enhances the performance of damage detection and localization.
STRUCTURAL HEALTH MONITORING-AN INTERNATIONAL JOURNAL
(2023)
Article
Physics, Multidisciplinary
Renze Dong, Hongze Leng, Juan Zhao, Junqiang Song, Shutian Liang
Summary: The initial field plays a crucial role in numerical weather prediction. Data assimilation, which is a method of obtaining the initial field for forecast models, is an effective approach. Machine learning has achieved success in various applications and can also be applied to operational weather prediction and data assimilation.
Article
Multidisciplinary Sciences
Claudia R. Arbeitman, Pablo Rojas, Pedro Ojeda-May, Martin E. Garcia
Summary: Research has shown that external electric fields can significantly destabilize the S protein of SARS-CoV-2, causing structural damage and disrupting its ability to bind to ACE2 receptors. This approach could offer a clean physical method to weaken the virus, without the need for further biochemical processing, for infection prevention and in-vitro structural manipulation purposes. The method's non-specificity suggests it could be applicable to other mutations in S, other proteins of SARS-CoV-2, and membrane proteins of other virus types.
NATURE COMMUNICATIONS
(2021)
Article
Mathematics, Applied
Luong Van Nguyen, Nguyen Thi Thu, Nguyen Thai An
Summary: In this paper, variational inequalities governed by strongly pseudomonotone vector fields on Hadamard manifolds are considered. The existence and uniqueness of the solution, linear convergence, error estimates, and finite convergence for sequences generated by a modified projection method for solving variational inequalities are investigated. Some examples and numerical experiments are also provided to illustrate the results.
APPLICABLE ANALYSIS
(2023)
Article
Chemistry, Medicinal
Subhamoy Mahajan, Tian Tang
Summary: This article presents a fully automated algorithm that can coarse-grain the molecular structure of polyethylenimine (PEI) from its all-atom simulation, allowing for replication of experimental data and computational inference of chemical structures. The developed methodology can be extended to other polymers.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Mathematics
Nazim Hussain Hajano, Muhammad Sabeel Khan, Lisheng Liu, Mumtaz Ali Kaloi, Hai Mei
Summary: In this paper, the monolithic Eulerian formulation is used to analyze the micro-structural effects of linearly increasing Reynolds number and mean inflow velocity on fluid flow. The obtained results show that the micro-rotational velocity field is significantly affected by the increase in these parameters.
Article
Chemistry, Physical
Hengli Zhao, Guillaume Maurin, Aziz Ghoufi
Summary: In this study, force-field molecular simulations were used to investigate the hexane isomer separation performance of the Zr channel-like metalorganic framework (MOF) MIL-140B under external mechanical pressure. It was found that applying a mechanical pressure of 1 GPa improved the quinary selectivity of the MOF by 80% compared to the mechanically unconstrained scenario. The attractive separation performance was attributed to the structural changes of MIL-140B, including linker tilting enhancement, pore size contraction, and conformational rearrangements of hexane isomers under mechanical pressure.
JOURNAL OF PHYSICAL CHEMISTRY C
(2022)
Article
Mathematics
Zheng Xu, Song Yan, Cong Wu, Qing Duan, Sixia Chen, Yun Li
Summary: This study developed NGS data-based methods for association studies, filling the gap in the literature. Simulation studies showed that NGS data-based methods have better performance than genotype-based methods for handling binary and count responses, especially when sequencing depth is low.
Article
Biochemistry & Molecular Biology
Erich R. Kuechler, Matthew Jacobson, Thibault Mayor, Jorg Gsponer
Summary: Phase separation-based condensate formation is a new working paradigm in biology that helps to understand cellular phenomena. GraPES is an online interface that provides propensity scores for protein localization in cellular condensates.
NUCLEIC ACIDS RESEARCH
(2022)
Article
Chemistry, Physical
Timothy J. Giese, Jinzhe Zeng, Solen Ekesan, Darrin M. York
Summary: This paper presents a fast, accurate, and robust approach for determining free energy profiles and kinetic isotope effects of RNA reactions with the inclusion of nuclear quantum effects. The approach utilizes deep potential range correction (DPRc) and quantum mechanical/molecular mechanical (QM/MM) simulations to achieve high accuracy in tuning QM and QM/MM interactions. The method is demonstrated through the calculation of free energy profiles of RNA cleavage reactions and reactions involving thio-substitutions, showing close agreement with ab initio calculations and the superior performance compared to other methods.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2022)
Article
Chemistry, Physical
Solen Ekesan, Erika McCarthy, David A. Case, Darrin M. York
Summary: Electrostatic interactions play a crucial role in the structure and function of RNA, particularly in the recruitment of metal ions for catalysis. This study investigates the electrostatic features and their relation to the binding of monovalent and divalent metal ions in metal-dependent ribozymes and an engineered DNAzyme. The findings demonstrate the importance of RNA electrostatics in catalysis, including the structural integrity of the active state, pK(a) tuning, and electrostatic stabilization of the transition state.
JOURNAL OF PHYSICAL CHEMISTRY B
(2022)
Article
Chemistry, Physical
Hsu-Chun Tsai, Tai-Sung Lee, Abir Ganguly, Timothy J. Giese, Maximilian C. C. J. C. Ebert, Paul Labute, Kenneth M. Merz, Darrin M. York
Summary: We propose a framework for optimized alchemical transformation pathways in free energy simulations using nonlinear mixing and a new functional form for softcore potentials. The framework is implemented and tested in the GPU-accelerated AMBER software suite. The optimized pathways integrate important features such as smoothstep functions, power scaling of interactions, LJ pairwise form, and smoothing of the potential at the nonbonded cutoff boundary. The pathways demonstrate superior numerical stability and minimal variance of free energy estimates compared to traditional methods.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Jinzhe Zeng, Yujun Tao, Timothy J. Giese, Darrin M. York
Summary: We introduce the QD7 pi-v1.0 model for accurately modeling the internal energy of drug molecules. This model combines a quantum mechanical/machine learning potential correction with a high-level deep-learning potential. It outperforms other semiempirical and machine learning potentials in handling electrostatic interactions and charge/protonation state changes. The QD pi model is highly accurate in various molecular interactions and shows excellent performance in relative protonation/deprotonation energies and tautomers.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Tai-Sung Lee, Hsu-Chun Tsai, Abir Ganguly, Darrin M. York
Summary: We propose an alchemical enhanced sampling method called ACES, implemented in the GPU-accelerated AMBER free energy MD engine. The method creates an enhanced sampling state by reducing or eliminating certain potential energy terms and interactions, while maintaining terms that limit the need for extensive phase space sampling. This enhanced sampling state is connected to the real state through a Hamiltonian replica exchange framework, resulting in a counterdiffusion of states. The ACES method has been successfully applied to various test cases and demonstrated superior performance compared to traditional MD and alternative enhanced sampling methods.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Multidisciplinary
Benjamin Weissman, Solen Ekesan, Hsuan-Chun Lin, Shahbaz Gardezi, Nan-Sheng Li, Timothy J. Giese, Erika McCarthy, Michael E. Harris, Darrin M. York, Joseph A. Piccirilli
Summary: Ribonucleases and small nucleolytic ribozymes both catalyze RNA strand cleavage through 2'-O-transphosphorylation, but their mechanisms involve distinct transition states. In this study, we demonstrate that hepatitis delta virus ribozyme catalysis proceeds through a dissociative, metaphosphate-like transition state, in contrast to the associative transition states observed with other enzymes. These findings provide evidence for a unique ribozyme active site design that modulates the RNA cleavage pathway.
JOURNAL OF THE AMERICAN CHEMICAL SOCIETY
(2023)
Article
Biochemistry & Molecular Biology
Suhyun Yoon, Edward Ollie, Darrin M. York, Joseph A. Piccirilli, Michael E. Harris
Summary: Psr is an important experimental system for defining RNA catalysis and designing valuable tools in biotechnology. The rate of Psr catalysis is too fast to measure manually and the reaction steps that limit catalysis are not well understood.
Article
Chemistry, Physical
Jinzhe Zeng, Yujun Tao, Timothy J. Giese, Darrin M. York
Summary: Modern semiempirical electronic structure methods have potential applications in drug discovery for accurately modeling biological and drug-like molecules. Comparisons were made between different methods, and the hybrid quantum mechanical/machine learning potentials, especially the QD pi model, showed the most robust performance for tautomers and protonation states relevant to drug discovery.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Timothy J. Giese, Darrin M. York
Summary: We used the modified Bigeleisen-Mayer equation to calculate kinetic isotope effect values for non-enzymatic phosphoryl transfer reactions. The modified equation includes the ratio of vibrational frequencies and the effect of isotopic substitution on the activation free energy. We developed a practical method to estimate the frequency ratio correction directly from umbrella sampling, which avoids the need for normal mode analysis. This method provides a new tool for calculating kinetic isotope effects in complex chemical reactions in the condensed phase.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Yuqing Xu, Michael E. Harris, Darrin M. York, Kin-Yiu Wong
Summary: RNA strand cleavage can be catalyzed by both ribozymes and hydroxide or hydronium ions. Experiments showed that cleavage of the 5'-linked nucleoside and isomerization between 3',5'- and 2',5'-phosphodiesters occur under acidic conditions, while only cleavage reaction is observed under basic conditions. A path-integral approach was used to reveal the reaction mechanisms under acidic conditions, and the proposed mechanisms can also be supported by experimental pH-rate profiles.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Biochemistry & Molecular Biology
Erika McCarthy, Soelen Ekesan, Timothy J. Giese, Timothy J. Wilson, Jie Deng, Lin Huang, David M. J. Lilley, Darrin M. York
Summary: By using simulations and calculations, we have elucidated the mechanism of methyl transfer catalyzed by a methyltransferase ribozyme. We have identified two transition states and a rate-controlling step, and predicted the activity-pH profile of the reaction.
NUCLEIC ACIDS RESEARCH
(2023)
Article
Biochemistry & Molecular Biology
Erich R. Kuechler, Matthew Jacobson, Thibault Mayor, Jorg Gsponer
Summary: Phase separation-based condensate formation is an important phenomenon in biology, and the GraPES web-server provides propensity scores for the localization of proteins into cellular condensates, which helps uncover the functional impact of these condensates.
NUCLEIC ACIDS RESEARCH
(2022)