Heqi Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Heqi Sun has a total of 48 co-authors. Here are the top co-authors by number of shared publications:
Hong Tan
Shanghai Jiao Tong University
6 shared publications- Dong-Qing Wei
School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, China
5 shared publications Junwei Chen
State Key Laboratory of Microbial Metabolism, Shanghai-Islamabad-Belgrade Joint Innovation Center on Antibacterial Resistances, Joint International Research Laboratory of Metabolic & Developmental Sciences, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, 200240, China
5 shared publicationsDong-Qing Wei
4 shared publicationsShenggeng Lin
State Key Laboratory of Microbial Metabolism, Shanghai-Islamabad-Belgrade Joint Innovation Center on Antibacterial Resistances, Joint International Research Laboratory of Metabolic & Developmental Sciences, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, 200240, China
4 shared publications- Yi Xiong
Shanghai Jiao Tong University
4 shared publications
Latest Publications
- Open Access
- 24 October 2026
- Published by Oxford University Press (OUP) in Genomics, Proteomics & Bioinformatics
RNASTOP: A Deep Learning Framework for mRNA Chemical Stability Prediction and Optimization
Messenger RNA (mRNA) vaccines offer promising therapeutic potential for combating various diseases, yet their inherent chemical instability hampers long-term efficacy. Although several methods have been developed to predict mRNA degradation, they often suffer from limited accuracy and lack the capability for rational codon optimization. Here, we propose RNASTOP, a novel framework that integrates deep learning with heuristic search to simultaneously predict and optimize mRNA chemical stability. RNASTOP achieves a 13% improvement in accuracy over the top-performing model on the Stanford OpenVaccine competition dataset and demonstrates robust generalization in predicting full-length mRNA degradation. When applied to mRNA codon optimization, RNASTOP reduces the minimum free energy of the varicella-zoster virus vaccine sequence by 75.73% while maintaining high translation efficiency. Overall, RNASTOP serves as a powerful tool for predicting and optimizing mRNA chemical stability, accelerating the development of mRNA therapeutics. The source code of RNASTOP is available at https://github.com/xlab-BioAI/RNASTOP.
- 24 October 2026
- Published by Oxford University Press (OUP) in Genomics, Proteomics & Bioinformatics
Open AccessRNASTOP: A Deep Learning Framework for mRNA Chemical Stability Prediction and Optimization
Messenger RNA (mRNA) vaccines offer promising therapeutic potential for combating various diseases, yet their inherent chemical instability hampers long-term efficacy. Although several methods have been developed to predict mRNA degradation, they often suffer from limited accuracy and lack the capability for rational codon optimization. Here, we propose RNASTOP, a novel framework that integrates deep learning with heuristic search to simultaneously predict and optimize mRNA chemical stability. RNASTOP achieves a 13% improvement in accuracy over the top-performing model on the Stanford OpenVaccine competition dataset and demonstrates robust generalization in predicting full-length mRNA degradation. When applied to mRNA codon optimization, RNASTOP reduces the minimum free energy of the varicella-zoster virus vaccine sequence by 75.73% while maintaining high translation efficiency. Overall, RNASTOP serves as a powerful tool for predicting and optimizing mRNA chemical stability, accelerating the development of mRNA therapeutics. The source code of RNASTOP is available at https://github.com/xlab-BioAI/RNASTOP.
- 28 August 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 66  (17) , 10806-10818
TvTPPIMI: A Boundary-Aware Dual-View Framework for Predicting PPI-Modulator Interactions
Protein–protein interactions (PPIs) regulate essential cellular processes and represent an important class of therapeutic targets; however, discovering effective modulators of PPIs remains a formidable challenge. Although deep learning approaches have been widely explored for PPI modulator discovery, many rely on simplified representations that obscure interchain boundary information and fine-grained PPI–modulator interaction (PPIMI) patterns, limiting their robustness under distribution shifts. To address this challenge, we introduce TvTPPIMI, a framework that leverages learnable boundary tokens to encode partner-aware boundary information and models PPIMI at atom–residue resolution. In a case study targeting the AURKA–TPX2 interaction, TvTPPIMI prioritized putative modulatory candidates from a small-molecule screening library. Structure-based docking, attention analysis, multireplica molecular dynamics simulations, MM/GBSA binding free-energy estimation, and noncovalent interaction analyses provided post hoc physical support for the stable AURKA binding of selected candidates, highlighting CE02–6266 as the most favorable compound among the tested hits. Together, these results suggest that TvTPPIMI provides a generalizable computational framework with coarse-grained, attention-based interpretive cues for PPIMI prediction and can be integrated with structure- and dynamics-based analyses to support PPI modulator discovery.
- 28 August 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 66  (17) , 10806-10818
TvTPPIMI: A Boundary-Aware Dual-View Framework for Predicting PPI-Modulator Interactions
Protein–protein interactions (PPIs) regulate essential cellular processes and represent an important class of therapeutic targets; however, discovering effective modulators of PPIs remains a formidable challenge. Although deep learning approaches have been widely explored for PPI modulator discovery, many rely on simplified representations that obscure interchain boundary information and fine-grained PPI–modulator interaction (PPIMI) patterns, limiting their robustness under distribution shifts. To address this challenge, we introduce TvTPPIMI, a framework that leverages learnable boundary tokens to encode partner-aware boundary information and models PPIMI at atom–residue resolution. In a case study targeting the AURKA–TPX2 interaction, TvTPPIMI prioritized putative modulatory candidates from a small-molecule screening library. Structure-based docking, attention analysis, multireplica molecular dynamics simulations, MM/GBSA binding free-energy estimation, and noncovalent interaction analyses provided post hoc physical support for the stable AURKA binding of selected candidates, highlighting CE02–6266 as the most favorable compound among the tested hits. Together, these results suggest that TvTPPIMI provides a generalizable computational framework with coarse-grained, attention-based interpretive cues for PPIMI prediction and can be integrated with structure- and dynamics-based analyses to support PPI modulator discovery.
- Open Access
- 9 April 2026
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 17  (20) , 10223-10236
Design of permeability-optimized target-binding macrocycles via direct preference optimization
CycDiff-DPO uses preference-aligned diffusion to design target-specific macrocyclic peptides with enhanced membrane permeability, balancing binding competence and cell permeability in one framework.
- 9 April 2026
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 17  (20) , 10223-10236
Open AccessDesign of permeability-optimized target-binding macrocycles via direct preference optimization
CycDiff-DPO uses preference-aligned diffusion to design target-specific macrocyclic peptides with enhanced membrane permeability, balancing binding competence and cell permeability in one framework.
- Open Access
RNASTOP: A Deep Learning Framework for mRNA Chemical Stability Prediction and Optimization
Messenger RNA (mRNA) vaccines offer promising therapeutics for combating various diseases, yet their inherent chemical instability hampers their long-term efficacy. Although several methods have been developed to predict mRNA degradation, they exhibit limited accuracy and lack the capability for rational sequence optimization. Here, we propose RNASTOP, a novel framework integrating deep learning with heuristic search to simultaneously predict and optimize mRNA chemical stability. RNASTOP achieves a 13% accuracy improvement over the top-performing model on the Stanford OpenVaccine competition dataset and demonstrates robust generalization in predicting full-length mRNA degradation. Applied to mRNA codon optimization, RNASTOP reduces the minimum free energy of the Varicella-Zoster Virus vaccine sequence by 75.73% while maintaining high translation efficiency. Overall, RNASTOP serves as a powerful tool for predicting and optimizing mRNA chemical stability, poised to expedite the development of mRNA therapeutics. The source code of RNASTOP can be accessed at https://github.com/xlab-BioAI/RNASTOP .
Open AccessRNASTOP: A Deep Learning Framework for mRNA Chemical Stability Prediction and Optimization
Messenger RNA (mRNA) vaccines offer promising therapeutics for combating various diseases, yet their inherent chemical instability hampers their long-term efficacy. Although several methods have been developed to predict mRNA degradation, they exhibit limited accuracy and lack the capability for rational sequence optimization. Here, we propose RNASTOP, a novel framework integrating deep learning with heuristic search to simultaneously predict and optimize mRNA chemical stability. RNASTOP achieves a 13% accuracy improvement over the top-performing model on the Stanford OpenVaccine competition dataset and demonstrates robust generalization in predicting full-length mRNA degradation. Applied to mRNA codon optimization, RNASTOP reduces the minimum free energy of the Varicella-Zoster Virus vaccine sequence by 75.73% while maintaining high translation efficiency. Overall, RNASTOP serves as a powerful tool for predicting and optimizing mRNA chemical stability, poised to expedite the development of mRNA therapeutics. The source code of RNASTOP can be accessed at https://github.com/xlab-BioAI/RNASTOP .
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30  (7) , 5569-5581
Prediction of Drug-Target Interactions Based on Hypergraph Neural Networks With Multimodal Feature Fusion
Accurate drug-target interaction prediction is vital for drug discovery and optimization. Traditional experimental methods, while effective, are time-intensive and costly. HyperGCN-DTI, a novel framework that explicitly advances beyond existing models such as CHL-DTI and HHDTI by leveraging hypergraph neural networks with a multimodal feature fusion strategy. While exsiting methods primarily focuses on low-order graph representations and fixed heterogeneous network structures, HyperGCN-DTI incorporates richer multimodal fused features including embeddings from pretrained language models and diverse biological networks and build robust hypergraphs that capture high-order multi-entity relationships within drug-target pairs. This dual-channel architecture effectively captures both local topological connections and higher-order structural dependencies. HyperGCN-DTI outperforms state-of-the-art DTI prediction models across multiple datasets and remains robust under imbalanced and large-scale real-world datasets, demonstrating its superior predictive power. The model demonstrates significant improvements when using multimodal features and hypergraph-based message passing, with sensitivity analysis confirming stability across hyperparameter variations. Top-ranked predictions are validated through biomedical literature and molecular docking, underscoring the reliability and practical relevance of our approach. HyperGCN-DTI is the first DTI prediction model to jointly integrate such a wide range of heterogeneous information sources with hypergraph representation, significantly enhancing accuracy and robustness, particularly in sparse or noisy settings. The proposed model offers a powerful and generalizable tool for accelerating drug development and target identification.
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30  (7) , 5569-5581
Prediction of Drug-Target Interactions Based on Hypergraph Neural Networks With Multimodal Feature Fusion
Accurate drug-target interaction prediction is vital for drug discovery and optimization. Traditional experimental methods, while effective, are time-intensive and costly. HyperGCN-DTI, a novel framework that explicitly advances beyond existing models such as CHL-DTI and HHDTI by leveraging hypergraph neural networks with a multimodal feature fusion strategy. While exsiting methods primarily focuses on low-order graph representations and fixed heterogeneous network structures, HyperGCN-DTI incorporates richer multimodal fused features including embeddings from pretrained language models and diverse biological networks and build robust hypergraphs that capture high-order multi-entity relationships within drug-target pairs. This dual-channel architecture effectively captures both local topological connections and higher-order structural dependencies. HyperGCN-DTI outperforms state-of-the-art DTI prediction models across multiple datasets and remains robust under imbalanced and large-scale real-world datasets, demonstrating its superior predictive power. The model demonstrates significant improvements when using multimodal features and hypergraph-based message passing, with sensitivity analysis confirming stability across hyperparameter variations. Top-ranked predictions are validated through biomedical literature and molecular docking, underscoring the reliability and practical relevance of our approach. HyperGCN-DTI is the first DTI prediction model to jointly integrate such a wide range of heterogeneous information sources with hypergraph representation, significantly enhancing accuracy and robustness, particularly in sparse or noisy settings. The proposed model offers a powerful and generalizable tool for accelerating drug development and target identification.
- 2 September 2025
- Published by Springer Nature in Interdisciplinary Sciences: Computational Life Sciences
- Vol. 18  (1) , 180-194
Hot-Spot-Guided Generative Deep Learning for Drug-Like PPI Inhibitor Design
Protein-protein interactions (PPIs) are essential therapeutic targets, yet their large and relatively flat interfaces hinder the development of small-molecule inhibitors. Traditional computational approaches rely heavily on existing chemical libraries or expert heuristics, restricting exploration of novel chemical space. To address these challenges, we present Hot2Mol, a generative deep learning framework for the de novo design of target-specific and drug-like PPI inhibitors. Hot2Mol captures crucial pharmacophoric features from hot-spot residues, allowing precise targeting of PPI interfaces while eliminating the need for known bioactive ligands. The framework integrates three main components: a conditional transformer for pharmacophore-guided, property-constrained molecular generation; an E(n)-equivariant graph neural network to ensure accurate spatial alignment with PPI hot-spot pharmacophores; a variational autoencoder to sample novel and diverse molecular structures. Comprehensive assessments demonstrate that Hot2Mol outperforms state-of-the-art models in binding affinity, drug-likeness, synthetic accessibility, novelty, and uniqueness. Molecular dynamics simulations further confirm the strong binding stability of generated compounds. Case studies underscore Hot2Mol's ability to design high-affinity and selective PPI inhibitors, highlighting its potential to accelerate rational PPI-targeted drug discovery.
- 2 September 2025
- Published by Springer Nature in Interdisciplinary Sciences: Computational Life Sciences
- Vol. 18  (1) , 180-194
Hot-Spot-Guided Generative Deep Learning for Drug-Like PPI Inhibitor Design
Protein-protein interactions (PPIs) are essential therapeutic targets, yet their large and relatively flat interfaces hinder the development of small-molecule inhibitors. Traditional computational approaches rely heavily on existing chemical libraries or expert heuristics, restricting exploration of novel chemical space. To address these challenges, we present Hot2Mol, a generative deep learning framework for the de novo design of target-specific and drug-like PPI inhibitors. Hot2Mol captures crucial pharmacophoric features from hot-spot residues, allowing precise targeting of PPI interfaces while eliminating the need for known bioactive ligands. The framework integrates three main components: a conditional transformer for pharmacophore-guided, property-constrained molecular generation; an E(n)-equivariant graph neural network to ensure accurate spatial alignment with PPI hot-spot pharmacophores; a variational autoencoder to sample novel and diverse molecular structures. Comprehensive assessments demonstrate that Hot2Mol outperforms state-of-the-art models in binding affinity, drug-likeness, synthetic accessibility, novelty, and uniqueness. Molecular dynamics simulations further confirm the strong binding stability of generated compounds. Case studies underscore Hot2Mol's ability to design high-affinity and selective PPI inhibitors, highlighting its potential to accelerate rational PPI-targeted drug discovery.
- Open Access
ProStab: Prediction of protein stability change upon mutations by protein language and inverse folding models
Predicting protein stability change upon mutation is critical for protein engineering, yet remains limited by the modeling assumptions of physics-based methods and the generalization bottlenecks of data-driven approaches. We present ProStab, a deep learning framework that integrates sequence- and structure-based information, including the mutation-aware sequence embeddings from protein language models and the geometric features extracted via an inverse folding model. Trained on the large-scale Megascale dataset, ProStab demonstrates strong performance across diverse test sets and robust generalization across distribution shifts between the training and test sets. In head-to-head comparisons, ProStab outperforms all state-of-the-art methods with consistently higher Spearman correlation and precision. To evaluate its practical utility, we experimentally validated ProStab-predicted mutations on the model enzyme transaminase. Among the 16 successfully expressed variants, 4 exhibited improved thermal stability. Remarkably, the 1st top-ranked predicted mutation yielded the highest observed enzymatic activity, retaining three-fold that of the wild type after 10 minutes at 40 °C. To facilitate broader application, a publicly accessible web server has been developed. We envisage that ProStab provides a scalable and accurate platform for intelligent protein stability design.
Open AccessProStab: Prediction of protein stability change upon mutations by protein language and inverse folding models
Predicting protein stability change upon mutation is critical for protein engineering, yet remains limited by the modeling assumptions of physics-based methods and the generalization bottlenecks of data-driven approaches. We present ProStab, a deep learning framework that integrates sequence- and structure-based information, including the mutation-aware sequence embeddings from protein language models and the geometric features extracted via an inverse folding model. Trained on the large-scale Megascale dataset, ProStab demonstrates strong performance across diverse test sets and robust generalization across distribution shifts between the training and test sets. In head-to-head comparisons, ProStab outperforms all state-of-the-art methods with consistently higher Spearman correlation and precision. To evaluate its practical utility, we experimentally validated ProStab-predicted mutations on the model enzyme transaminase. Among the 16 successfully expressed variants, 4 exhibited improved thermal stability. Remarkably, the 1st top-ranked predicted mutation yielded the highest observed enzymatic activity, retaining three-fold that of the wild type after 10 minutes at 40 °C. To facilitate broader application, a publicly accessible web server has been developed. We envisage that ProStab provides a scalable and accurate platform for intelligent protein stability design.
- Open Access
Target-specific design of drug-like PPI inhibitors via hotspot-guided generative deep learning
Protein–protein interactions (PPIs) are vital therapeutic targets. However, the large and flat PPI interfaces pose challenges for the development of small-molecule inhibitors. Traditional computer-aided drug design approaches typically rely on pre-existing libraries or expert knowledge, limiting the exploration of novel chemical spaces needed for effective PPI inhibition. To overcome these limitations, we introduce Hot2Mol, a deep learning framework for the de novo design of drug-like, target-specific PPI inhibitors. Hot2Mol generates small molecules by mimicking the pharmacophoric features of hot-spot residues, enabling precise targeting of PPI interfaces without the need for bioactive ligands. The framework integrates three key components: a conditional transformer for pharmacophore-guided, drug-likeness-constrained molecular generation; an E(n)-equivariant graph neural network for accurate alignment with PPI hot-spot pharmacophores; and a variational autoencoder for generating novel and diverse molecular structures. Experimental evaluations demonstrate that Hot2Mol outperforms baseline models across multiple metrics, including docking affinities, drug-likenesses, synthetic accessibility, validity, uniqueness, and novelty. Furthermore, molecular dynamics simulations confirm the good binding stability of the generated molecules. Case studies underscore Hot2Mol’s ability to design high-affinity and selective PPI inhibitors, demonstrating its potential to accelerate rational PPI drug discovery.
Open AccessTarget-specific design of drug-like PPI inhibitors via hotspot-guided generative deep learning
Protein–protein interactions (PPIs) are vital therapeutic targets. However, the large and flat PPI interfaces pose challenges for the development of small-molecule inhibitors. Traditional computer-aided drug design approaches typically rely on pre-existing libraries or expert knowledge, limiting the exploration of novel chemical spaces needed for effective PPI inhibition. To overcome these limitations, we introduce Hot2Mol, a deep learning framework for the de novo design of drug-like, target-specific PPI inhibitors. Hot2Mol generates small molecules by mimicking the pharmacophoric features of hot-spot residues, enabling precise targeting of PPI interfaces without the need for bioactive ligands. The framework integrates three key components: a conditional transformer for pharmacophore-guided, drug-likeness-constrained molecular generation; an E(n)-equivariant graph neural network for accurate alignment with PPI hot-spot pharmacophores; and a variational autoencoder for generating novel and diverse molecular structures. Experimental evaluations demonstrate that Hot2Mol outperforms baseline models across multiple metrics, including docking affinities, drug-likenesses, synthetic accessibility, validity, uniqueness, and novelty. Furthermore, molecular dynamics simulations confirm the good binding stability of the generated molecules. Case studies underscore Hot2Mol’s ability to design high-affinity and selective PPI inhibitors, demonstrating its potential to accelerate rational PPI drug discovery.
- 1 December 2023
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 63  (23) , 7363-7372
A Multimodal Deep Learning Framework for Predicting PPI-Modulator Interactions
Protein-protein interactions (PPIs) are essential for various biological processes and diseases. However, most existing computational methods for identifying PPI modulators require either target structure or reference modulators, which restricts their applicability to novel PPI targets. To address this challenge, we propose MultiPPIMI, a sequence-based deep learning framework that predicts the interaction between any given PPI target and modulator. MultiPPIMI integrates multimodal representations of PPI targets and modulators and uses a bilinear attention network to capture intermolecular interactions. Experimental results on our curated benchmark data set show that MultiPPIMI achieves an average AUROC of 0.837 in three cold-start scenarios and an AUROC of 0.994 in the random-split scenario. Furthermore, the case study shows that MultiPPIMI can assist molecular docking simulations in screening inhibitors of Keap1/Nrf2 PPI interactions. We believe that the proposed method provides a promising way to screen PPI-targeted modulators.
- 1 December 2023
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 63  (23) , 7363-7372
A Multimodal Deep Learning Framework for Predicting PPI-Modulator Interactions
Protein-protein interactions (PPIs) are essential for various biological processes and diseases. However, most existing computational methods for identifying PPI modulators require either target structure or reference modulators, which restricts their applicability to novel PPI targets. To address this challenge, we propose MultiPPIMI, a sequence-based deep learning framework that predicts the interaction between any given PPI target and modulator. MultiPPIMI integrates multimodal representations of PPI targets and modulators and uses a bilinear attention network to capture intermolecular interactions. Experimental results on our curated benchmark data set show that MultiPPIMI achieves an average AUROC of 0.837 in three cold-start scenarios and an AUROC of 0.994 in the random-split scenario. Furthermore, the case study shows that MultiPPIMI can assist molecular docking simulations in screening inhibitors of Keap1/Nrf2 PPI interactions. We believe that the proposed method provides a promising way to screen PPI-targeted modulators.
- Open Access
Interface-aware molecular generative framework for protein-protein interaction modulators
Protein-protein interactions (PPIs) play a crucial role in numerous biochemical and biological processes. Although several structure-based molecular generative models have been developed, PPI interfaces and compounds targeting PPIs exhibit distinct physicochemical properties compared to traditional binding pockets and small-molecule drugs. As a result, generating compounds that effectively target PPIs, particularly by considering PPI complexes or interface hotspot residues, remains a significant challenge. In this work, we constructed a comprehensive dataset of PPI interfaces with active and inactive compound pairs. Based on this, we propose a novel molecular generative framework tailored to PPI interfaces, named GENiPPI. Our evaluation demonstrates that GENiPPI captures the implicit relationships between the PPI interfaces and the active molecules, and can generate novel compounds that target these interfaces. Moreover, GENiPPI can generate structurally diverse novel compounds with limited PPI interface modulators. To the best of our knowledge, this is the first exploration of a structure-based molecular generative model focused on PPI interfaces, which could facilitate the design of PPI modulators. The PPI interface-based molecular generative model enriches the existing landscape of structure-based (pocket/interface) molecular generative model.
Open AccessInterface-aware molecular generative framework for protein-protein interaction modulators
Protein-protein interactions (PPIs) play a crucial role in numerous biochemical and biological processes. Although several structure-based molecular generative models have been developed, PPI interfaces and compounds targeting PPIs exhibit distinct physicochemical properties compared to traditional binding pockets and small-molecule drugs. As a result, generating compounds that effectively target PPIs, particularly by considering PPI complexes or interface hotspot residues, remains a significant challenge. In this work, we constructed a comprehensive dataset of PPI interfaces with active and inactive compound pairs. Based on this, we propose a novel molecular generative framework tailored to PPI interfaces, named GENiPPI. Our evaluation demonstrates that GENiPPI captures the implicit relationships between the PPI interfaces and the active molecules, and can generate novel compounds that target these interfaces. Moreover, GENiPPI can generate structurally diverse novel compounds with limited PPI interface modulators. To the best of our knowledge, this is the first exploration of a structure-based molecular generative model focused on PPI interfaces, which could facilitate the design of PPI modulators. The PPI interface-based molecular generative model enriches the existing landscape of structure-based (pocket/interface) molecular generative model.