Yi Isaac Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yi Isaac Yang has a total of 77 co-authors. Here are the top co-authors by number of shared publications:
- Yi Qin Gao
Peking University
26 shared publications - Jun Zhang
Changping Laboratory
15 shared publications Lijiang Yang
College of Chemistry and Molecular Engineering
13 shared publicationsJun Zhang
Changping Laboratory 102200 Beijing China
11 shared publicationsYao-Kun Lei
Institute of Systems and Physical Biology, Shenzhen Bay Laboratory
8 shared publicationsZhen Zhang
School of Physics and Technology, Tangshan Normal University, 063000 Tangshan, China
8 shared publications
Latest Publications
- Open Access
- 14 October 2026
- Published by Wiley in Advanced Science
- p. e78094
Bridging Chemical Literature and Molecular Graphs: LLM‐Augmented Enantioselectivity Prediction for Iridium‐Catalyzed Asymmetric C–H Borylation
Asymmetric C–H borylation provides direct access to enantioenriched organoboron compounds. Predictive catalyst design for this reaction class remains challenging, as stereocontrol arises from coupled interactions among catalyst, substrate, and product structures. Moreover, mechanistic knowledge is dispersed across the literature. Here we report a machine‐learning framework for predicting enantioselectivity in iridium‐catalyzed asymmetric C–H borylation that bridges molecular graphs and chemical literature. A unified dataset comprising 800 literature‐reported reactions was constructed for this reaction class. The model combines a multi‐entity three‐dimensional graph encoder for catalyst, substrate, and product with temperature‐aware kinetic treatment, and a large‐language‐model prior module that extracts atom‐resolved mechanistic annotations from primary publications. On a scaffold‐based held‐out test set, the STEREO‐Lit (Spatial Three‐Entity Reaction Encoder Literature) model achieves strong predictive performance, outperforming conventional machine‐learning baselines and a geometry‐grounded control. Perturbation studies further show that literature‐derived semantic annotations contribute chemically meaningful, site‐specific information rather than merely increasing model capacity. The trained model was subsequently applied to explore a reaction space of about 100 000 catalyst–substrate combinations and identify promising high‐improvement candidates, several of which were supported by DFT calculations. The framework couples multi‐entity 3D encoding with literature‐derived priors, and remains predictive when training data are limited to a few hundred reactions.
- 14 October 2026
- Published by Wiley in Advanced Science
- p. e78094
Open AccessBridging Chemical Literature and Molecular Graphs: LLM‐Augmented Enantioselectivity Prediction for Iridium‐Catalyzed Asymmetric C–H Borylation
Asymmetric C–H borylation provides direct access to enantioenriched organoboron compounds. Predictive catalyst design for this reaction class remains challenging, as stereocontrol arises from coupled interactions among catalyst, substrate, and product structures. Moreover, mechanistic knowledge is dispersed across the literature. Here we report a machine‐learning framework for predicting enantioselectivity in iridium‐catalyzed asymmetric C–H borylation that bridges molecular graphs and chemical literature. A unified dataset comprising 800 literature‐reported reactions was constructed for this reaction class. The model combines a multi‐entity three‐dimensional graph encoder for catalyst, substrate, and product with temperature‐aware kinetic treatment, and a large‐language‐model prior module that extracts atom‐resolved mechanistic annotations from primary publications. On a scaffold‐based held‐out test set, the STEREO‐Lit (Spatial Three‐Entity Reaction Encoder Literature) model achieves strong predictive performance, outperforming conventional machine‐learning baselines and a geometry‐grounded control. Perturbation studies further show that literature‐derived semantic annotations contribute chemically meaningful, site‐specific information rather than merely increasing model capacity. The trained model was subsequently applied to explore a reaction space of about 100 000 catalyst–substrate combinations and identify promising high‐improvement candidates, several of which were supported by DFT calculations. The framework couples multi‐entity 3D encoding with literature‐derived priors, and remains predictive when training data are limited to a few hundred reactions.
- 10 January 2026
- Published by Springer Nature
- p. 239-288
Harnessing Artificial Intelligence to Revolutionise Molecular Modelling and Simulations
- 10 January 2026
- Published by Springer Nature
- p. 239-288
Harnessing Artificial Intelligence to Revolutionise Molecular Modelling and Simulations
- Open Access
Collective Variables Based on Multipole Expansion of Ewald Summation for Crystallization
Crystallization, a fundamental phase transition process governing material formation in natural and industrial contexts, involves the spontaneous emergence of long-range structural order from disordered phases. This long-range periodicity involves spatial and molecular orientation order. Molecular dynamics (MD) simulations of crystallization require collective variables (CVs) that accurately distinguish this long-\range periodicity. Existing CVs based on local descriptors (e.g., bond-orientational order) often lack transferability across crystal structures. To address this, we propose a unified CV framework derived from the multipole expansion of Ewald summation: a mathematical formalism bridging X-ray diffraction (XRD) principles and electrostatic energy computation in MD. By projecting atomic configurations onto a basis of spherical harmonics (complete for angular function representation), our CV achieves high-fidelity encoding of both translational and orientational order. Metadynamics simulations demonstrate that this CV drives efficient sampling of polymorphic pathways for known crystals and predicts stable phases even without crystal structures. This approach shows potential as a transferable platform for ab initio crystal structure prediction.
Open AccessCollective Variables Based on Multipole Expansion of Ewald Summation for Crystallization
Crystallization, a fundamental phase transition process governing material formation in natural and industrial contexts, involves the spontaneous emergence of long-range structural order from disordered phases. This long-range periodicity involves spatial and molecular orientation order. Molecular dynamics (MD) simulations of crystallization require collective variables (CVs) that accurately distinguish this long-\range periodicity. Existing CVs based on local descriptors (e.g., bond-orientational order) often lack transferability across crystal structures. To address this, we propose a unified CV framework derived from the multipole expansion of Ewald summation: a mathematical formalism bridging X-ray diffraction (XRD) principles and electrostatic energy computation in MD. By projecting atomic configurations onto a basis of spherical harmonics (complete for angular function representation), our CV achieves high-fidelity encoding of both translational and orientational order. Metadynamics simulations demonstrate that this CV drives efficient sampling of polymorphic pathways for known crystals and predicts stable phases even without crystal structures. This approach shows potential as a transferable platform for ab initio crystal structure prediction.
- 30 July 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 21 (15) , 7279-7289
Performing Path Integral Molecular Dynamics Using an Artificial Intelligence-Enhanced Molecular Simulation Framework
This study employed an artificial intelligence-enhanced molecular simulation framework to enable efficient path integral molecular dynamics (PIMD) simulations. Owing to its modular architecture and high-throughput capabilities, the framework effectively mitigates the computational complexity and resource-intensive limitations associated with conventional PIMD approaches. By integrating machine learning force fields (MLFFs) into the framework, we rigorously tested its performance through two representative cases: a small-molecule reaction system (double-proton transfer in the formic acid dimer) and a bulk-phase transition system (water-ice phase transformation). Computational results demonstrate that the proposed framework achieves accelerated PIMD simulations while preserving the quantum mechanical accuracy. These findings show that nuclear quantum effects can be captured for complex molecular systems using relatively low computational cost.
- 30 July 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 21 (15) , 7279-7289
Performing Path Integral Molecular Dynamics Using an Artificial Intelligence-Enhanced Molecular Simulation Framework
This study employed an artificial intelligence-enhanced molecular simulation framework to enable efficient path integral molecular dynamics (PIMD) simulations. Owing to its modular architecture and high-throughput capabilities, the framework effectively mitigates the computational complexity and resource-intensive limitations associated with conventional PIMD approaches. By integrating machine learning force fields (MLFFs) into the framework, we rigorously tested its performance through two representative cases: a small-molecule reaction system (double-proton transfer in the formic acid dimer) and a bulk-phase transition system (water-ice phase transformation). Computational results demonstrate that the proposed framework achieves accelerated PIMD simulations while preserving the quantum mechanical accuracy. These findings show that nuclear quantum effects can be captured for complex molecular systems using relatively low computational cost.
- Open Access
- 2 June 2025
- Published by American Chemical Society (ACS) in JACS Au
- Vol. 5 (6) , 2898-2908
A Sinking Approach to Explore Arbitrary Areas in Free Energy Landscapes
To address the time-scale limitations in molecular dynamics (MD) simulations, numerous enhanced sampling methods have been developed to expedite the exploration of complex free energy landscapes. A commonly employed approach accelerates the sampling of degrees of freedom associated with predefined collective variables (CVs), which typically tend to traverse the entire CV range. However, in many scenarios, the focus of interest is on specific regions within the CV space. In this paper, we introduce a novel “sinking” approach that enables enhanced sampling of arbitrary areas within the CV space. This method, referred to as SinkMeta, “sinks” the interior bias potential to create a restraining potential “cliff” at the grid edges, thus confining the exploration of CVs in MD simulations to a predefined area. SinkMeta requires minimal sampling steps to estimate the free energy landscape for CV subspaces of various shapes and dimensions, offering an efficient and flexible solution for sampling minimum free energy paths in high-dimensional spaces. We believe that SinkMeta will pioneer a new paradigm for sampling partial phase spaces and provide an efficient and straightforward way to study the interaction of drugs with biomolecules such as proteins and DNA in MD simulations.
- 2 June 2025
- Published by American Chemical Society (ACS) in JACS Au
- Vol. 5 (6) , 2898-2908
Open AccessA Sinking Approach to Explore Arbitrary Areas in Free Energy Landscapes
To address the time-scale limitations in molecular dynamics (MD) simulations, numerous enhanced sampling methods have been developed to expedite the exploration of complex free energy landscapes. A commonly employed approach accelerates the sampling of degrees of freedom associated with predefined collective variables (CVs), which typically tend to traverse the entire CV range. However, in many scenarios, the focus of interest is on specific regions within the CV space. In this paper, we introduce a novel “sinking” approach that enables enhanced sampling of arbitrary areas within the CV space. This method, referred to as SinkMeta, “sinks” the interior bias potential to create a restraining potential “cliff” at the grid edges, thus confining the exploration of CVs in MD simulations to a predefined area. SinkMeta requires minimal sampling steps to estimate the free energy landscape for CV subspaces of various shapes and dimensions, offering an efficient and flexible solution for sampling minimum free energy paths in high-dimensional spaces. We believe that SinkMeta will pioneer a new paradigm for sampling partial phase spaces and provide an efficient and straightforward way to study the interaction of drugs with biomolecules such as proteins and DNA in MD simulations.
- Open Access
Performing Path Integral Molecular Dynamics Using Artificial Intelligence Enhanced Molecular Simulation Framework
This study employed an artificial intelligence-enhanced molecular simulation framework to enable efficient Path Integral Molecular Dynamics (PIMD) simulations. Owing to its modular architecture and high-throughput capabilities, the framework effectively mitigates the computational complexity and resource-intensive limitations associated with conventional PIMD approaches. By integrating machine learning force fields (MLFFs) into the framework, we rigorously tested its performance through two representative cases: a small-molecule reaction system (double proton transfer in formic acid dimer) and a bulk-phase transition system (water-ice phase transformation). Computational results demonstrate that the proposed framework achieves accelerated PIMD simulations while preserving quantum mechanical accuracy. These findings show that nuclear quantum effects can be captured for complex molecular systems, using relatively low computational cost.
Open AccessPerforming Path Integral Molecular Dynamics Using Artificial Intelligence Enhanced Molecular Simulation Framework
This study employed an artificial intelligence-enhanced molecular simulation framework to enable efficient Path Integral Molecular Dynamics (PIMD) simulations. Owing to its modular architecture and high-throughput capabilities, the framework effectively mitigates the computational complexity and resource-intensive limitations associated with conventional PIMD approaches. By integrating machine learning force fields (MLFFs) into the framework, we rigorously tested its performance through two representative cases: a small-molecule reaction system (double proton transfer in formic acid dimer) and a bulk-phase transition system (water-ice phase transformation). Computational results demonstrate that the proposed framework achieves accelerated PIMD simulations while preserving quantum mechanical accuracy. These findings show that nuclear quantum effects can be captured for complex molecular systems, using relatively low computational cost.
- 27 February 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 21 (5) , 2501-2514
Multiscale Force Field Model Based on a Graph Neural Network for Complex Chemical Systems
Inspired by the QM/MM methodology, the ML/MM approach introduces a new opportunity for multiscale simulation, improving the balance between accuracy and computational efficiency. Benefited from the rapid advancements in molecular embedding methods, density functional theory level quantum mechanical (QM) calculations within the QM/MM framework can be accelerated by several orders of magnitude through the application of machine learning (ML) potential energy surfaces. As a problem inherited from the QM/MM methodology, challenges exist in designing the interactions between machine learning and molecular mechanics (MM) regions. In this study, electrostatic interactions between machine learning and MM atoms are treated by using a graphical neural network based on stationary perturbation theory. In this protocol, we process coordinates and MM charges to yield electrostatic energy and forces, resulting in a high-performance electrostatic embedding ML/MM architecture. The accuracy of the ML/MM energy was validated in aqueous solutions of alanine dipeptide and allyl vinyl ether (AVE). We investigated the transferability of parameters trained from AVE in a single solvent to various other solvents, including water, methanol, dimethyl sulfoxide, toluene, ionic liquids, and water-toluene interface environments. We then established a solvent-free protocol for data set preparation. Comparison of the free energy landscapes of the Claisen rearrangement of AVE in different solvation environments showed the catalytic effect of aqueous solutions, consistent with experiments.
- 27 February 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 21 (5) , 2501-2514
Multiscale Force Field Model Based on a Graph Neural Network for Complex Chemical Systems
Inspired by the QM/MM methodology, the ML/MM approach introduces a new opportunity for multiscale simulation, improving the balance between accuracy and computational efficiency. Benefited from the rapid advancements in molecular embedding methods, density functional theory level quantum mechanical (QM) calculations within the QM/MM framework can be accelerated by several orders of magnitude through the application of machine learning (ML) potential energy surfaces. As a problem inherited from the QM/MM methodology, challenges exist in designing the interactions between machine learning and molecular mechanics (MM) regions. In this study, electrostatic interactions between machine learning and MM atoms are treated by using a graphical neural network based on stationary perturbation theory. In this protocol, we process coordinates and MM charges to yield electrostatic energy and forces, resulting in a high-performance electrostatic embedding ML/MM architecture. The accuracy of the ML/MM energy was validated in aqueous solutions of alanine dipeptide and allyl vinyl ether (AVE). We investigated the transferability of parameters trained from AVE in a single solvent to various other solvents, including water, methanol, dimethyl sulfoxide, toluene, ionic liquids, and water-toluene interface environments. We then established a solvent-free protocol for data set preparation. Comparison of the free energy landscapes of the Claisen rearrangement of AVE in different solvation environments showed the catalytic effect of aqueous solutions, consistent with experiments.
- 12 February 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 21 (4) , 1990-1996
Effective Nucleation Size for Ice Crystallization
Despite the apparent simplicity of water molecules, the kinetics of ice nucleation under natural conditions can be surprisingly intricate. Previous studies have yielded critical nucleation sizes that vary widely due to differences in experimental and computational approaches. In our investigation, we employed all-atom molecular dynamics simulations to explore spontaneously grown and ideal ice nuclei, revealing significant disparities in their kinetics. Notably, nucleation defects challenge the applicability of the classical nucleation theory (CNT) to spontaneously grown ice nuclei. To address this, we propose a generalized nucleation theory that effectively describes the kinetics of ice crystal nucleation across diverse conditions. The kinetics of ice nuclei, as characterized by the "corrected" critical nucleus size, follow a linear law akin to that assumed by CNT. This generalized nucleation theory also provides insights for studying the kinetics of other crystalline materials.
- 12 February 2025
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 21 (4) , 1990-1996
Effective Nucleation Size for Ice Crystallization
Despite the apparent simplicity of water molecules, the kinetics of ice nucleation under natural conditions can be surprisingly intricate. Previous studies have yielded critical nucleation sizes that vary widely due to differences in experimental and computational approaches. In our investigation, we employed all-atom molecular dynamics simulations to explore spontaneously grown and ideal ice nuclei, revealing significant disparities in their kinetics. Notably, nucleation defects challenge the applicability of the classical nucleation theory (CNT) to spontaneously grown ice nuclei. To address this, we propose a generalized nucleation theory that effectively describes the kinetics of ice crystal nucleation across diverse conditions. The kinetics of ice nuclei, as characterized by the "corrected" critical nucleus size, follow a linear law akin to that assumed by CNT. This generalized nucleation theory also provides insights for studying the kinetics of other crystalline materials.
- 4 July 2024
- Published by Wiley in WIREs Computational Molecular Science
- Vol. 14 (4)
From predicting to decision making: Reinforcement learning in biomedicine
Reinforcement learning (RL) is one important branch of artificial intelligence (AI), which intuitively imitates the learning style of human beings. It is commonly derived from solving game playing problems and is extensively used for decision‐making, control and optimization problems. It has been extensively applied for solving complicated problems with the property of Markov decision‐making processes. With data accumulation and comprehensive analysis, researchers are not only satisfied with predicting the results for experimental systems but also hope to design or control them for the sake of obtaining the desired properties or functions. RL is potentially facilitated to solve a large number of complicated biological and chemical problems, because they could be decomposed into multi‐step decision‐making process. In practice, substantial progress has been made in the application of RL to the field of biomedicine. In this paper, we will first give a brief description about RL, including its definition, basic theory and different type of methods. Then we will review some detailed applications in various domains, for example, molecular design, reaction planning, molecular simulation and etc. In the end, we will summarize the essentialities of RL approaches to solve more diverse problems compared with other machine learning methods and also outlook the possible trends to overcome their limitations in the future.This article is categorized under: Data Science > Chemoinformatics Data Science > Computer Algorithms and Programming Data Science > Artificial Intelligence/Machine Learning
- 4 July 2024
- Published by Wiley in WIREs Computational Molecular Science
- Vol. 14 (4)
From predicting to decision making: Reinforcement learning in biomedicine
Reinforcement learning (RL) is one important branch of artificial intelligence (AI), which intuitively imitates the learning style of human beings. It is commonly derived from solving game playing problems and is extensively used for decision‐making, control and optimization problems. It has been extensively applied for solving complicated problems with the property of Markov decision‐making processes. With data accumulation and comprehensive analysis, researchers are not only satisfied with predicting the results for experimental systems but also hope to design or control them for the sake of obtaining the desired properties or functions. RL is potentially facilitated to solve a large number of complicated biological and chemical problems, because they could be decomposed into multi‐step decision‐making process. In practice, substantial progress has been made in the application of RL to the field of biomedicine. In this paper, we will first give a brief description about RL, including its definition, basic theory and different type of methods. Then we will review some detailed applications in various domains, for example, molecular design, reaction planning, molecular simulation and etc. In the end, we will summarize the essentialities of RL approaches to solve more diverse problems compared with other machine learning methods and also outlook the possible trends to overcome their limitations in the future.This article is categorized under: Data Science > Chemoinformatics Data Science > Computer Algorithms and Programming Data Science > Artificial Intelligence/Machine Learning
- 17 May 2024
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry A
- Vol. 128 (21) , 4378-4390
Generating High-Precision Force Fields for Molecular Dynamics Simulations to Study Chemical Reaction Mechanisms Using Molecular Configuration Transformer
Theoretical studies on chemical reaction mechanisms have been crucial in organic chemistry. Traditionally, calculating the manually constructed molecular conformations of transition states for chemical reactions using quantum chemical calculations is the most commonly used method. However, this way is heavily dependent on individual experience and chemical intuition. In our previous study, we proposed a research paradigm that used enhanced sampling in molecular dynamics simulations to study chemical reactions. This approach can directly simulate the entire process of a chemical reaction. However, the computational speed limited the use of high-precision potential energy functions for simulations. To address this issue, we presented a scheme for training high-precision force fields for molecular modeling using a previously developed graph-neural-network-based molecular model, molecular configuration transformer. This potential energy function allowed for highly accurate simulations at a low computational cost, leading to more precise calculations of the mechanism of chemical reactions. We applied this approach to study a Claisen rearrangement reaction and a carbonyl insertion reaction catalyzed by manganese.
- 17 May 2024
- Published by American Chemical Society (ACS) in The Journal of Physical Chemistry A
- Vol. 128 (21) , 4378-4390
Generating High-Precision Force Fields for Molecular Dynamics Simulations to Study Chemical Reaction Mechanisms Using Molecular Configuration Transformer
Theoretical studies on chemical reaction mechanisms have been crucial in organic chemistry. Traditionally, calculating the manually constructed molecular conformations of transition states for chemical reactions using quantum chemical calculations is the most commonly used method. However, this way is heavily dependent on individual experience and chemical intuition. In our previous study, we proposed a research paradigm that used enhanced sampling in molecular dynamics simulations to study chemical reactions. This approach can directly simulate the entire process of a chemical reaction. However, the computational speed limited the use of high-precision potential energy functions for simulations. To address this issue, we presented a scheme for training high-precision force fields for molecular modeling using a previously developed graph-neural-network-based molecular model, molecular configuration transformer. This potential energy function allowed for highly accurate simulations at a low computational cost, leading to more precise calculations of the mechanism of chemical reactions. We applied this approach to study a Claisen rearrangement reaction and a carbonyl insertion reaction catalyzed by manganese.