Yurong Zou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Yurong Zou has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
- Lijuan Chen
Sichuan University
2 shared publications Zhongning Guo
West China Hospital, Sichuan University
2 shared publicationsBaozhu Fang
School of Computer Science, Nanjing Normal University, Nanjing, China
1 shared publication- Bingke Li
School of Biology and Chemical Engineering/Henan Key Laboratory of Microbial Fermentation, Nanyang Institute of Technology, Henan, Nanyang 473000, PR China
1 shared publication Bo Chen
School of Computer Science, Nanjing Normal University, Nanjing, China
1 shared publicationChenghua Zhang
School of Basic Medical Sciences North Sichuan Medical College Nanchong 637007 P. R. China
1 shared publication
Latest Publications
- Open Access
- 9 April 2026
- Published by Springer Nature in Research Square
ArcMol Enables Task-Adaptive Spherical Representation Learning for Molecular Property Prediction
Molecular properties are measured under different experimental and representational settings, each capturing a distinct aspect of molecular behavior. Integrating such task-dependent molecular information remains challenging, especially when labeled data are scarce and structurally similar molecules exhibit markedly different properties. Here we introduce ArcMol, a task-adaptive framework for molecular property prediction that combines explicit physicochemical descriptors with learned molecular representations. ArcMol organizes these features in a hyperspherical latent space, where molecular similarity is defined by angular relationships. Across more than 150 molecular property prediction benchmarks, ArcMol achieves performance comparable to state-of-the-art methods, while showing improved robustness in low-data settings and for structurally similar molecules. Analyses of the learned representations reveal task-relevant separation, ordered property gradients, and coherent local neighborhoods. We demonstrate ArcMol in a real-world application targeting the kinase IRAK4, where it prioritizes candidate molecules for drug repurposing and leads to the identification of nanomolar type II inhibitors. These results suggest ArcMol as a general framework for molecular property prediction in discovery settings.
- 9 April 2026
- Published by Springer Nature in Research Square
Open AccessArcMol Enables Task-Adaptive Spherical Representation Learning for Molecular Property Prediction
Molecular properties are measured under different experimental and representational settings, each capturing a distinct aspect of molecular behavior. Integrating such task-dependent molecular information remains challenging, especially when labeled data are scarce and structurally similar molecules exhibit markedly different properties. Here we introduce ArcMol, a task-adaptive framework for molecular property prediction that combines explicit physicochemical descriptors with learned molecular representations. ArcMol organizes these features in a hyperspherical latent space, where molecular similarity is defined by angular relationships. Across more than 150 molecular property prediction benchmarks, ArcMol achieves performance comparable to state-of-the-art methods, while showing improved robustness in low-data settings and for structurally similar molecules. Analyses of the learned representations reveal task-relevant separation, ordered property gradients, and coherent local neighborhoods. We demonstrate ArcMol in a real-world application targeting the kinase IRAK4, where it prioritizes candidate molecules for drug repurposing and leads to the identification of nanomolar type II inhibitors. These results suggest ArcMol as a general framework for molecular property prediction in discovery settings.
- Open Access
- 12 March 2025
- Published by Springer Nature in Communications Biology
- Vol. 8 (1) , 1-15
A structure-based framework for selective inhibitor design and optimization
Structure-based drug design aims to create active compounds with favorable properties by analyzing target structures. Recently, deep generative models have facilitated structure-specific molecular generation. However, many methods are limited by inadequate pharmaceutical data, resulting in suboptimal molecular properties and unstable conformations. Additionally, these approaches often overlook binding pocket interactions and struggle with selective inhibitor design. To address these challenges, we developed a framework called Coarse-grained and Multi-dimensional Data-driven molecular generation (CMD-GEN). CMD-GEN bridges ligand-protein complexes with drug-like molecules by utilizing coarse-grained pharmacophore points sampled from diffusion model, enriching training data. Through a hierarchical architecture, it decomposes three-dimensional molecule generation within the pocket into pharmacophore point sampling, chemical structure generation, and conformation alignment, mitigating instability issues. CMD-GEN outperforms other methods in benchmark tests and controls drug-likeness effectively. Furthermore, CMD-GEN excels in cases across three synthetic lethal targets, and wet-lab validation with PARP1/2 inhibitors confirms its potential in selective inhibitor design.
- 12 March 2025
- Published by Springer Nature in Communications Biology
- Vol. 8 (1) , 1-15
Open AccessA structure-based framework for selective inhibitor design and optimization
Structure-based drug design aims to create active compounds with favorable properties by analyzing target structures. Recently, deep generative models have facilitated structure-specific molecular generation. However, many methods are limited by inadequate pharmaceutical data, resulting in suboptimal molecular properties and unstable conformations. Additionally, these approaches often overlook binding pocket interactions and struggle with selective inhibitor design. To address these challenges, we developed a framework called Coarse-grained and Multi-dimensional Data-driven molecular generation (CMD-GEN). CMD-GEN bridges ligand-protein complexes with drug-like molecules by utilizing coarse-grained pharmacophore points sampled from diffusion model, enriching training data. Through a hierarchical architecture, it decomposes three-dimensional molecule generation within the pocket into pharmacophore point sampling, chemical structure generation, and conformation alignment, mitigating instability issues. CMD-GEN outperforms other methods in benchmark tests and controls drug-likeness effectively. Furthermore, CMD-GEN excels in cases across three synthetic lethal targets, and wet-lab validation with PARP1/2 inhibitors confirms its potential in selective inhibitor design.
- 23 September 2020
- Published by Wiley in ChemistrySelect
- Vol. 5 (36) , 11112-11119
Virtual Screening and Discovery of Matrix Metalloproteinase‐12 Inhibitors by Swarm Intelligence Optimization Algorithm‐Based Machine Learning
Matrix metalloproteinase‐12 (MMP‐12) is an attractive therapeutic target for drug design and discovery for many human conditions. In this study, six swarm intelligence optimization algorithms were applied to optimize the parameters of the model generated using the LibSVM toolkit in MATLAB to identify potential MMP‐12 inhibitors (MMP‐12is); six types of optimized support vector machine (SVM) models were established. The highest prediction accuracy obtained was 98.89 %, which was equivalent to the effect of the optimal “RF+opt” model. All six models passed the Y‐randomization test and showed excellent performance with reliable results. Virtual screening identified 371 molecules with a predictive probability score greater than 0.9. The optimized SVM models, in addition to “RF+opt” and “SVM2” models, were combined to establish a consistency evaluation system. Our results revealed six non‐toxic potential MMP‐12is. This process provides a strong theoretical basis for the design, synthesis, and development of novel drugs targeting MMP‐12.
- 23 September 2020
- Published by Wiley in ChemistrySelect
- Vol. 5 (36) , 11112-11119
Virtual Screening and Discovery of Matrix Metalloproteinase‐12 Inhibitors by Swarm Intelligence Optimization Algorithm‐Based Machine Learning
Matrix metalloproteinase‐12 (MMP‐12) is an attractive therapeutic target for drug design and discovery for many human conditions. In this study, six swarm intelligence optimization algorithms were applied to optimize the parameters of the model generated using the LibSVM toolkit in MATLAB to identify potential MMP‐12 inhibitors (MMP‐12is); six types of optimized support vector machine (SVM) models were established. The highest prediction accuracy obtained was 98.89 %, which was equivalent to the effect of the optimal “RF+opt” model. All six models passed the Y‐randomization test and showed excellent performance with reliable results. Virtual screening identified 371 molecules with a predictive probability score greater than 0.9. The optimized SVM models, in addition to “RF+opt” and “SVM2” models, were combined to establish a consistency evaluation system. Our results revealed six non‐toxic potential MMP‐12is. This process provides a strong theoretical basis for the design, synthesis, and development of novel drugs targeting MMP‐12.
- 1 August 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 838-843
A Deep Learning-Oriented Personal Credit Model for Online Learning Platform
There are several problems in online learning platform, such as the lack of effective supervision of learning process and the low completion rate, the general situation of slack learning attitude and cheating, the lack of scientific planning of learning methods and the serious fragmentation, and the low standard-reaching rate of learning results and the poor credibility. In order to solve these problems, this paper defines the concept of personal credit in learning, and declares the characteristics of personal credit from the four aspects of online learning process, attitude, method and results based on the comparative analysis of deep learning and surface learning in educational theory, then builds a personal credit model oriented to deep learning in online learning platform. The model includes five evaluation dimensions of participation, positivity, regularity, qualified rate and influence, as well as 14 explicit or implicit behavior indicators, especially the corresponding learning behavior abnormality detection mechanism. The experimental results show that this model has the characteristics of scientific clustering index, comprehensive evaluation angles and reliable exception handling, which can effectively evaluate learners' learning behavior state and provide a new way for online learning platform to implement educational contracts and share services.
- 1 August 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 838-843
A Deep Learning-Oriented Personal Credit Model for Online Learning Platform
There are several problems in online learning platform, such as the lack of effective supervision of learning process and the low completion rate, the general situation of slack learning attitude and cheating, the lack of scientific planning of learning methods and the serious fragmentation, and the low standard-reaching rate of learning results and the poor credibility. In order to solve these problems, this paper defines the concept of personal credit in learning, and declares the characteristics of personal credit from the four aspects of online learning process, attitude, method and results based on the comparative analysis of deep learning and surface learning in educational theory, then builds a personal credit model oriented to deep learning in online learning platform. The model includes five evaluation dimensions of participation, positivity, regularity, qualified rate and influence, as well as 14 explicit or implicit behavior indicators, especially the corresponding learning behavior abnormality detection mechanism. The experimental results show that this model has the characteristics of scientific clustering index, comprehensive evaluation angles and reliable exception handling, which can effectively evaluate learners' learning behavior state and provide a new way for online learning platform to implement educational contracts and share services.