Yanjing Wang
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
Yanjing Wang has a total of 158 co-authors. Here are the top co-authors by number of shared publications:
- Dong-Qing Wei
School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, China
23 shared publications - Buyong Ma
Engineering Research Center of Cell & Therapeutic Antibody (MOE), School of Pharmacy, Shanghai Jiao Tong University, Shanghai, China
10 shared publications Abbas Khan
Engineering Research Center of Cell & Therapeutic Antibody, School of Pharmacy, Shanghai Jiao Tong University, Shanghai, 200240, China; Department of Bioinformatics and Biological Statistics, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, 200240, PR China
9 shared publications- Xiangeng Wang
State Key Laboratory of Microbial Metabolism, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, China
6 shared publications - Yi Xiong
Shanghai Jiao Tong University
6 shared publications Dong-Qing Wei
School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, 200030, China
5 shared publications
Latest Publications
- Open Access
- 23 September 2026
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 27 (19) , 8500
SynerHyper: Predicting Anticancer Drug Synergy for Lung Cancer Using Uni-Mol Enhanced Hypergraph Neural Networks and Multimodal Fusion
Lung cancer remains a leading cause of cancer-related mortality worldwide, and combination therapy is an important strategy for improving treatment efficacy and overcoming drug resistance. However, identifying synergistic drug combinations remains challenging because of the large combinatorial search space and the strong dependence of drug responses on cellular context. Here, we present SynerHyper, a multimodal computational framework for lung cancer drug synergy prediction that integrates three-dimensional molecular representations, transcriptomic profiles, hypergraph learning, and ensemble fusion. Drug molecules are encoded using a Uni-Mol-based molecular encoder, while high-dimensional cell-line gene-expression profiles are compressed using a deep autoencoder. Each drug–drug–cell-line observation is represented as a three-node hyperedge, enabling a two-layer hypergraph neural network to model higher-order interactions among the two drugs and their cellular context. Data augmentation is applied exclusively to the training subset to increase training-data diversity and mitigate class imbalance, and predictions from the hypergraph model, multilayer perceptron, and LightGBM are combined through soft ensemble fusion. Under the split-before-augmentation evaluation protocol, the validation set was used for model selection, while the test set remained completely isolated until final evaluation. SynerHyper achieved a validation AUC of 0.815 and an independent test AUC of 0.808, outperforming the strongest baseline DeepDDS (GCN, 0.792) and other representative methods including AttenSyn (0.780), DeepSynergy (0.750), and DTSyn (0.690) under the same lung cancer dataset. Together with an accessible online prediction server (SynerHyper), SynerHyper provides a lung cancer-specific computational framework for prioritizing candidate drug combinations for downstream experimental validation.
- 23 September 2026
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 27 (19) , 8500
Open AccessSynerHyper: Predicting Anticancer Drug Synergy for Lung Cancer Using Uni-Mol Enhanced Hypergraph Neural Networks and Multimodal Fusion
Lung cancer remains a leading cause of cancer-related mortality worldwide, and combination therapy is an important strategy for improving treatment efficacy and overcoming drug resistance. However, identifying synergistic drug combinations remains challenging because of the large combinatorial search space and the strong dependence of drug responses on cellular context. Here, we present SynerHyper, a multimodal computational framework for lung cancer drug synergy prediction that integrates three-dimensional molecular representations, transcriptomic profiles, hypergraph learning, and ensemble fusion. Drug molecules are encoded using a Uni-Mol-based molecular encoder, while high-dimensional cell-line gene-expression profiles are compressed using a deep autoencoder. Each drug–drug–cell-line observation is represented as a three-node hyperedge, enabling a two-layer hypergraph neural network to model higher-order interactions among the two drugs and their cellular context. Data augmentation is applied exclusively to the training subset to increase training-data diversity and mitigate class imbalance, and predictions from the hypergraph model, multilayer perceptron, and LightGBM are combined through soft ensemble fusion. Under the split-before-augmentation evaluation protocol, the validation set was used for model selection, while the test set remained completely isolated until final evaluation. SynerHyper achieved a validation AUC of 0.815 and an independent test AUC of 0.808, outperforming the strongest baseline DeepDDS (GCN, 0.792) and other representative methods including AttenSyn (0.780), DeepSynergy (0.750), and DTSyn (0.690) under the same lung cancer dataset. Together with an accessible online prediction server (SynerHyper), SynerHyper provides a lung cancer-specific computational framework for prioritizing candidate drug combinations for downstream experimental validation.
- Open Access
CXCL14 Inhibits Colon Cancer Progression by Modulating Tumor Cell Invasion and Immune Microenvironment
CXCL14 is a highly conserved chemokine with potential roles in tumor progression and immune modulation. This study investigates the functional impact of CXCL14 on colon cancer by exploring its effects on tumor cell behavior and the immune microenvironment. We generated stable cell lines overexpressing CXCL14 in mouse MC38 and CT26 cells and human HCT15 colon cancer cells, and used these models to assess tumor growth, invasion, and immune cell infiltration. Our results demonstrate that CXCL14 suppresses colon cancer cell proliferation, migration, and metastasis. In vitro, CXCL14 inhibited the expression of matrix metalloproteinases (MMPs), key regulators of epithelial–mesenchymal transition (EMT), suggesting a role in promoting mesenchymal–epithelial transition (MET). Additionally, in vivo studies using a subcutaneous tumor model showed that CXCL14 not only suppressed tumor growth but also enhanced the infiltration of immune cells, including NK cells, dendritic cells (DCs), and T cells, converting the tumor microenvironment from a “cold” to a “hot” phenotype. RNA sequencing and pathway analyses revealed that CXCL14 regulates the expression of genes associated with angiogenesis, immune response, and cell signaling, particularly through the MAPK pathway. Furthermore, CXCL14’s influence on tumor progression was confirmed in a spleen-to-liver metastasis model, where its overexpression reduced metastatic spread. In conclusion, CXCL14 inhibits colon cancer progression by modulating both tumor cell behavior and the immune landscape, making it a promising candidate for targeted immunotherapy. Our findings highlight CXCL14’s potential to enhance anti-tumor immunity and provide new insights into its therapeutic applications in colon cancer.
Open AccessCXCL14 Inhibits Colon Cancer Progression by Modulating Tumor Cell Invasion and Immune Microenvironment
CXCL14 is a highly conserved chemokine with potential roles in tumor progression and immune modulation. This study investigates the functional impact of CXCL14 on colon cancer by exploring its effects on tumor cell behavior and the immune microenvironment. We generated stable cell lines overexpressing CXCL14 in mouse MC38 and CT26 cells and human HCT15 colon cancer cells, and used these models to assess tumor growth, invasion, and immune cell infiltration. Our results demonstrate that CXCL14 suppresses colon cancer cell proliferation, migration, and metastasis. In vitro, CXCL14 inhibited the expression of matrix metalloproteinases (MMPs), key regulators of epithelial–mesenchymal transition (EMT), suggesting a role in promoting mesenchymal–epithelial transition (MET). Additionally, in vivo studies using a subcutaneous tumor model showed that CXCL14 not only suppressed tumor growth but also enhanced the infiltration of immune cells, including NK cells, dendritic cells (DCs), and T cells, converting the tumor microenvironment from a “cold” to a “hot” phenotype. RNA sequencing and pathway analyses revealed that CXCL14 regulates the expression of genes associated with angiogenesis, immune response, and cell signaling, particularly through the MAPK pathway. Furthermore, CXCL14’s influence on tumor progression was confirmed in a spleen-to-liver metastasis model, where its overexpression reduced metastatic spread. In conclusion, CXCL14 inhibits colon cancer progression by modulating both tumor cell behavior and the immune landscape, making it a promising candidate for targeted immunotherapy. Our findings highlight CXCL14’s potential to enhance anti-tumor immunity and provide new insights into its therapeutic applications in colon cancer.
- Open Access
- 31 August 2025
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 26 (5)
SageTCR: a structure-based model integrating residue- and atom-level representations for enhanced TCR-pMHC binding prediction
T-cell receptors (TCRs) recognize peptide–MHC (pMHC) complexes through intricate structural interactions, which is a core component of adaptive immunity. However, the diverse and cross-reactive nature of TCRs poses great challenges for accurate prediction of TCR-epitope interactions, hampering the advancement and broad application of TCR-related therapies. Here, we present SageTCR, a bi-level graph neural network (GNN) framework that leverages structural data to predict TCR-pMHC binding possibilities. Harnessing the pretrained language models, SageTCR encodes detailed structural arrangement at both residue-level and atomic-level and effectively integrates the bimodal representations via attention mechanisms. To tackle the deficiency of experimental structures, we explore comprehensive data augmentation strategies to enrich the training and increase the generalizability while concurrently preserving the characteristic TCR-pMHC diagonal binding mode. SageTCR demonstrates superior performance compared to six methods with different deep learning architectures. Furthermore, SageTCR offers the interpretability by identifying and focusing on the conformational features of pivotal contact residues on the interface, which can provide valuable insights for TCR engineering and immunotherapy design.
- 31 August 2025
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 26 (5)
Open AccessSageTCR: a structure-based model integrating residue- and atom-level representations for enhanced TCR-pMHC binding prediction
T-cell receptors (TCRs) recognize peptide–MHC (pMHC) complexes through intricate structural interactions, which is a core component of adaptive immunity. However, the diverse and cross-reactive nature of TCRs poses great challenges for accurate prediction of TCR-epitope interactions, hampering the advancement and broad application of TCR-related therapies. Here, we present SageTCR, a bi-level graph neural network (GNN) framework that leverages structural data to predict TCR-pMHC binding possibilities. Harnessing the pretrained language models, SageTCR encodes detailed structural arrangement at both residue-level and atomic-level and effectively integrates the bimodal representations via attention mechanisms. To tackle the deficiency of experimental structures, we explore comprehensive data augmentation strategies to enrich the training and increase the generalizability while concurrently preserving the characteristic TCR-pMHC diagonal binding mode. SageTCR demonstrates superior performance compared to six methods with different deep learning architectures. Furthermore, SageTCR offers the interpretability by identifying and focusing on the conformational features of pivotal contact residues on the interface, which can provide valuable insights for TCR engineering and immunotherapy design.
- Open Access
- 12 August 2025
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 16 (38) , 17885-17899
SAGERank: inductive learning of protein–protein interaction from antibody–antigen recognition
Predicting Antibody–Antigen (Ab–Ag) docking and structure-based design represent significant long-term and therapeutically important challenges in computational biology. We present SAGERank, a general, configurable deep learning framework for antibody design using Graph Sample and Aggregate Networks. SAGERank successfully predicted the majority of epitopes in a cancer target dataset. In nanobody–antigen structure prediction, SAGERank, coupled with a protein dynamics structure prediction algorithm, slightly outperforms Alphafold3. Most importantly, our study demonstrates the real potential of inductive deep learning to overcome the small dataset problem in molecular science. The SAGERank models trained for antibody–antigen docking can be used to examine general protein–protein interaction tasks, such as T Cell Receptor-peptide-Major Histocompatibility Complex (TCR-pMHC) recognition, classification of biological versus crystal interfaces, and prediction of ternary complexes of molecular glues. In the cases of ranking docking decoys and identifying biological interfaces, SAGERank is competitive with or outperforms state-of-the-art methods.
- 12 August 2025
- Published by Royal Society of Chemistry (RSC) in Chemical Science
- Vol. 16 (38) , 17885-17899
Open AccessSAGERank: inductive learning of protein–protein interaction from antibody–antigen recognition
Predicting Antibody–Antigen (Ab–Ag) docking and structure-based design represent significant long-term and therapeutically important challenges in computational biology. We present SAGERank, a general, configurable deep learning framework for antibody design using Graph Sample and Aggregate Networks. SAGERank successfully predicted the majority of epitopes in a cancer target dataset. In nanobody–antigen structure prediction, SAGERank, coupled with a protein dynamics structure prediction algorithm, slightly outperforms Alphafold3. Most importantly, our study demonstrates the real potential of inductive deep learning to overcome the small dataset problem in molecular science. The SAGERank models trained for antibody–antigen docking can be used to examine general protein–protein interaction tasks, such as T Cell Receptor-peptide-Major Histocompatibility Complex (TCR-pMHC) recognition, classification of biological versus crystal interfaces, and prediction of ternary complexes of molecular glues. In the cases of ranking docking decoys and identifying biological interfaces, SAGERank is competitive with or outperforms state-of-the-art methods.
- Open Access
- 10 April 2025
- Published by Elsevier in Journal of Pharmaceutical Analysis
- Vol. 15 (6) , 101303
Scaffold and SAR studies on c-MET inhibitors using machine learning approaches
- 10 April 2025
- Published by Elsevier in Journal of Pharmaceutical Analysis
- Vol. 15 (6) , 101303
Open AccessScaffold and SAR studies on c-MET inhibitors using machine learning approaches
- Open Access
- 9 October 2024
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 25 (19) , 10857
Insights of Expression Profile of Chemokine Family in Inflammatory Bowel Diseases and Carcinogenesis
Chemokines are integral components of the immune system and deeply involved in the pathogenesis and progression of inflammatory bowel disease (IBD) and colorectal cancer (CRC). Although a considerable amount of transcriptome data has been accumulated on these diseases, most of them are limited to a specific stage of the disease. The purpose of this study is to visually demonstrate the dynamic changes in chemokines across various stages of bowel diseases by integrating relevant datasets. Integrating the existing datasets for IBD and CRC, we compare the expression changes of chemokines across different pathological stages. This study collected 11 clinical databases from various medical centers around the world. Patients: Data of patient tissue types were classified into IBD, colorectal adenoma, primary carcinoma, metastasis, and healthy control according to the publisher’s annotation. The expression changes in chemokines in various pathological stages are statistically analyzed. The chemokines were clustered by different expression patterns. The chemokine family was clustered into four distinct expression patterns, which correspond to varying expression changes in different stages of colitis and tumor development. Certain chemokines and receptors associated with inflammation and tumorigenesis have been identified. Furthermore, it was confirmed that the 2,4,6-trinitrobenzenesulfonic acid (TNBS)-induced colitis model and the azoxymethane (AOM)/ dextran sulfate sodium (DSS)-induced colon cancer model shows stronger correlations with the clinical data in terms of chemokine expression levels. This study paints a panoramic picture of the expression profiles of chemokine families at multiple stages from IBD to advanced colon cancer, facilitating a comprehensive understanding of the regulation patterns of chemokines and guiding the direction of drug development. This study provides researchers with a clear atlas of chemokine expression in the pathological processes of inflammatory bowel disease and colon cancer.
- 9 October 2024
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 25 (19) , 10857
Open AccessInsights of Expression Profile of Chemokine Family in Inflammatory Bowel Diseases and Carcinogenesis
Chemokines are integral components of the immune system and deeply involved in the pathogenesis and progression of inflammatory bowel disease (IBD) and colorectal cancer (CRC). Although a considerable amount of transcriptome data has been accumulated on these diseases, most of them are limited to a specific stage of the disease. The purpose of this study is to visually demonstrate the dynamic changes in chemokines across various stages of bowel diseases by integrating relevant datasets. Integrating the existing datasets for IBD and CRC, we compare the expression changes of chemokines across different pathological stages. This study collected 11 clinical databases from various medical centers around the world. Patients: Data of patient tissue types were classified into IBD, colorectal adenoma, primary carcinoma, metastasis, and healthy control according to the publisher’s annotation. The expression changes in chemokines in various pathological stages are statistically analyzed. The chemokines were clustered by different expression patterns. The chemokine family was clustered into four distinct expression patterns, which correspond to varying expression changes in different stages of colitis and tumor development. Certain chemokines and receptors associated with inflammation and tumorigenesis have been identified. Furthermore, it was confirmed that the 2,4,6-trinitrobenzenesulfonic acid (TNBS)-induced colitis model and the azoxymethane (AOM)/ dextran sulfate sodium (DSS)-induced colon cancer model shows stronger correlations with the clinical data in terms of chemokine expression levels. This study paints a panoramic picture of the expression profiles of chemokine families at multiple stages from IBD to advanced colon cancer, facilitating a comprehensive understanding of the regulation patterns of chemokines and guiding the direction of drug development. This study provides researchers with a clear atlas of chemokine expression in the pathological processes of inflammatory bowel disease and colon cancer.
- Open Access
- 2 December 2023
- Published by MDPI AG in Biomolecules
- Vol. 13 (12) , 1735
Advances in Drug Design and Development for Human Therapeutics Using Artificial Intelligence-II
Building on our 2021–2022 Special Issue, “Advances in Drug Design and Development for Human Therapeutics Using Artificial Intelligence
- 2 December 2023
- Published by MDPI AG in Biomolecules
- Vol. 13 (12) , 1735
Open AccessAdvances in Drug Design and Development for Human Therapeutics Using Artificial Intelligence-II
Building on our 2021–2022 Special Issue, “Advances in Drug Design and Development for Human Therapeutics Using Artificial Intelligence
- Open Access
- 7 November 2023
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 24 (22) , 16027
Data Mining Suggests That CXCL14 Gene Silencing in Colon Cancer Is Due to Promoter Methylation
CXCL14 is one of the most evolutionarily conserved members of the chemokine family and is constitutionally expressed in multiple organs, suggesting that it is involved in the homeostasis maintenance of the system. CXCL14 is highly expressed in colon epithelial cells and shows obvious gene silencing in clinical colon cancer samples, suggesting that its silencing is related to the immune escape of cancer cells. In this paper, we analyzed the expression profiles of multiple human clinical colon cancer datasets and mouse colon cancer models to reveal the variation trend of CXCL14 expression during colitis, colon polyps, primary colon cancer, and liver metastases. The relationship between CXCL14 gene silencing and promoter hypermethylation was revealed through the colorectal carcinoma methylation database. The results suggest that CXCL14 is a tumor suppressor gene in colorectal carcinoma which is activated first and then silenced during the process of tumor occurrence and deterioration. Promoter hypermethylation is the main cause of CXCL14 silencing. The methylation level of CXCL14 is correlated with the anatomic site of tumor occurrence, positively correlated with patient age, and associated with prognosis. Reversing the hypermethylation of CXCL14 may be an epigenetic therapy for colon cancer.
- 7 November 2023
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 24 (22) , 16027
Open AccessData Mining Suggests That CXCL14 Gene Silencing in Colon Cancer Is Due to Promoter Methylation
CXCL14 is one of the most evolutionarily conserved members of the chemokine family and is constitutionally expressed in multiple organs, suggesting that it is involved in the homeostasis maintenance of the system. CXCL14 is highly expressed in colon epithelial cells and shows obvious gene silencing in clinical colon cancer samples, suggesting that its silencing is related to the immune escape of cancer cells. In this paper, we analyzed the expression profiles of multiple human clinical colon cancer datasets and mouse colon cancer models to reveal the variation trend of CXCL14 expression during colitis, colon polyps, primary colon cancer, and liver metastases. The relationship between CXCL14 gene silencing and promoter hypermethylation was revealed through the colorectal carcinoma methylation database. The results suggest that CXCL14 is a tumor suppressor gene in colorectal carcinoma which is activated first and then silenced during the process of tumor occurrence and deterioration. Promoter hypermethylation is the main cause of CXCL14 silencing. The methylation level of CXCL14 is correlated with the anatomic site of tumor occurrence, positively correlated with patient age, and associated with prognosis. Reversing the hypermethylation of CXCL14 may be an epigenetic therapy for colon cancer.
- Open Access
- 2 November 2023
- Published by Springer Nature in BMC Genomics
- Vol. 24 (1) , 1-12
Exploring microproteins from various model organisms using the mip-mining database
Microproteins, prevalent across all kingdoms of life, play a crucial role in cell physiology and human health. Although global gene transcription is widely explored and abundantly available, our understanding of microprotein functions using transcriptome data is still limited. To mitigate this problem, we present a database, Mip-mining (https://weilab.sjtu.edu.cn/mipmining/), underpinned by high-quality RNA-sequencing data exclusively aimed at analyzing microprotein functions. The Mip-mining hosts 336 sets of high-quality transcriptome data from 8626 samples and nine representative living organisms, including microorganisms, plants, animals, and humans, in our Mip-mining database. Our database specifically provides a focus on a range of diseases and environmental stress conditions, taking into account chemical, physical, biological, and diseases-related stresses. Comparatively, our platform enables customized analysis by inputting desired data sets with self-determined cutoff values. The practicality of Mip-mining is demonstrated by identifying essential microproteins in different species and revealing the importance of ATP15 in the acetic acid stress tolerance of budding yeast. We believe that Mip-mining will facilitate a greater understanding and application of microproteins in biotechnology. Moreover, it will be beneficial for designing therapeutic strategies under various biological conditions.
- 2 November 2023
- Published by Springer Nature in BMC Genomics
- Vol. 24 (1) , 1-12
Open AccessExploring microproteins from various model organisms using the mip-mining database
Microproteins, prevalent across all kingdoms of life, play a crucial role in cell physiology and human health. Although global gene transcription is widely explored and abundantly available, our understanding of microprotein functions using transcriptome data is still limited. To mitigate this problem, we present a database, Mip-mining (https://weilab.sjtu.edu.cn/mipmining/), underpinned by high-quality RNA-sequencing data exclusively aimed at analyzing microprotein functions. The Mip-mining hosts 336 sets of high-quality transcriptome data from 8626 samples and nine representative living organisms, including microorganisms, plants, animals, and humans, in our Mip-mining database. Our database specifically provides a focus on a range of diseases and environmental stress conditions, taking into account chemical, physical, biological, and diseases-related stresses. Comparatively, our platform enables customized analysis by inputting desired data sets with self-determined cutoff values. The practicality of Mip-mining is demonstrated by identifying essential microproteins in different species and revealing the importance of ATP15 in the acetic acid stress tolerance of budding yeast. We believe that Mip-mining will facilitate a greater understanding and application of microproteins in biotechnology. Moreover, it will be beneficial for designing therapeutic strategies under various biological conditions.
- Open Access
SAGERank: Inductive Learning of Protein-Protein Interaction from Antibody-Antigen Recognition using Graph Sample and Aggregate Networks Framework
Numerous experiments and computationally solved antibody-antigen interfaces offer the possibility of training deep-learning models to help predict their biological correlations. Predicting antibody-antigen docking and structure-based design represent significant long-term and therapeutically important challenges in computational biology. We present SAGERank, a general, configurable deep learning framework for antibody design using Graph Sample and Aggregate Networks, which mainly includes ranking docking decoys, detecting binding, and identifying biological interfaces. The model proved its reliability in three different tasks. For both problems ranking docking decoys and identifying biological interfaces, SAGERank is competitive with or outperforms, state-of-the-art methods. Besides, the SAGERank model still showed a high degree of confidence in determining whether the antibody-antigen could bind. All of these demonstrate the versatility of SAGERank for structural biology research. Most importantly, our study demonstrated the real potential of inductive deep learning to overcome small dataset problem in molecular science. The SAGERank models trained for antibody-antigen docking can be used to examine generally protein-protein interaction docking and differentiate crystal packing from biological interface.
Open AccessSAGERank: Inductive Learning of Protein-Protein Interaction from Antibody-Antigen Recognition using Graph Sample and Aggregate Networks Framework
Numerous experiments and computationally solved antibody-antigen interfaces offer the possibility of training deep-learning models to help predict their biological correlations. Predicting antibody-antigen docking and structure-based design represent significant long-term and therapeutically important challenges in computational biology. We present SAGERank, a general, configurable deep learning framework for antibody design using Graph Sample and Aggregate Networks, which mainly includes ranking docking decoys, detecting binding, and identifying biological interfaces. The model proved its reliability in three different tasks. For both problems ranking docking decoys and identifying biological interfaces, SAGERank is competitive with or outperforms, state-of-the-art methods. Besides, the SAGERank model still showed a high degree of confidence in determining whether the antibody-antigen could bind. All of these demonstrate the versatility of SAGERank for structural biology research. Most importantly, our study demonstrated the real potential of inductive deep learning to overcome small dataset problem in molecular science. The SAGERank models trained for antibody-antigen docking can be used to examine generally protein-protein interaction docking and differentiate crystal packing from biological interface.