Siliang Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Siliang Chen has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
- Gu He
Sichuan University and Collaborative Innovation Center for Biotherapy
2 shared publications - Yaoxi Su
Department of Dermatology, West China Hospital, Sichuan University, Chengdu, China
2 shared publications Aixian Zhang
Department of Dermatology and Venereology, West China Hospital, Sichuan University, Chengdu 610041, China; The Fifth Affiliated Hospital (The Fifth Clinical Medical College), Xinjiang Medical University, Urumqi 830011, China
1 shared publication- Bin Wang
School of Pharmacy
1 shared publication Daihan Wang
Laboratory of Dermatology Clinical Institute of Inflammation and Immunology, Frontiers Science Center for Disease‐Related Molecular Network, State Key Laboratory of Biotherapy, West China Hospital, Sichuan University Chengdu China
1 shared publication- Dan Du
Department of Dermatology & Venereology, West China Hospital, Sichuan University, Chengdu 610041, China
1 shared publication
Latest Publications
- 6 September 2026
- Published by Wiley in WIREs Computational Molecular Science
- Vol. 16 (5) , e70086
EGNN ‐Based Generative Models for 3D Molecular Generation
Equivariant graph neural networks (EGNNs) are becoming the geometric infrastructure of 3D molecular generation for AI‐aided drug discovery. By enforcing E(n), E(3), or SE(3) equivariance, they separate physical molecular structure from arbitrary coordinate‐frame conventions and ensure that predicted coordinates, denoising directions, and velocity fields transform consistently with molecular geometry. This symmetry‐aware design supports direct modeling of conformations and protein–ligand spatial relationships while remaining compatible with diffusion, flow‐based, autoregressive, and hybrid generative frameworks. Consequently, EGNNs enable more geometrically consistent generation and facilitate controllable design conditioned on binding pockets, pharmacophores, fragments, scaffolds, reference ligands, or molecular properties. Their benefits, however, are bounded by what equivariance encodes. Coordinate‐frame consistency does not itself enforce valid bonds, valence, stereochemistry, topology–geometry agreement, synthesizability, or biological activity. Moreover, practical models must reconcile discrete chemical variables with continuous geometric dynamics, represent long‐range interactions without prohibitive computational cost, and account for protein flexibility, solvent, metal ions, and induced fit. Performance is further constrained by biased docking or property predictors, heterogeneous benchmarks, and limited prospective validation. This review synthesizes how EGNNs function across major generative paradigms and argues that future drug‐oriented systems should combine joint 2D–3D graph generation, physically credible interaction modeling, synthesis and ADMET constraints, multi‐objective optimization, and closed‐loop experimental feedback. EGNNs should therefore be viewed not as a complete chemical solution, but as the symmetry‐aware foundation upon which more testable and biologically relevant molecular design systems can be built.
- 6 September 2026
- Published by Wiley in WIREs Computational Molecular Science
- Vol. 16 (5) , e70086
EGNN ‐Based Generative Models for 3D Molecular Generation
Equivariant graph neural networks (EGNNs) are becoming the geometric infrastructure of 3D molecular generation for AI‐aided drug discovery. By enforcing E(n), E(3), or SE(3) equivariance, they separate physical molecular structure from arbitrary coordinate‐frame conventions and ensure that predicted coordinates, denoising directions, and velocity fields transform consistently with molecular geometry. This symmetry‐aware design supports direct modeling of conformations and protein–ligand spatial relationships while remaining compatible with diffusion, flow‐based, autoregressive, and hybrid generative frameworks. Consequently, EGNNs enable more geometrically consistent generation and facilitate controllable design conditioned on binding pockets, pharmacophores, fragments, scaffolds, reference ligands, or molecular properties. Their benefits, however, are bounded by what equivariance encodes. Coordinate‐frame consistency does not itself enforce valid bonds, valence, stereochemistry, topology–geometry agreement, synthesizability, or biological activity. Moreover, practical models must reconcile discrete chemical variables with continuous geometric dynamics, represent long‐range interactions without prohibitive computational cost, and account for protein flexibility, solvent, metal ions, and induced fit. Performance is further constrained by biased docking or property predictors, heterogeneous benchmarks, and limited prospective validation. This review synthesizes how EGNNs function across major generative paradigms and argues that future drug‐oriented systems should combine joint 2D–3D graph generation, physically credible interaction modeling, synthesis and ADMET constraints, multi‐objective optimization, and closed‐loop experimental feedback. EGNNs should therefore be viewed not as a complete chemical solution, but as the symmetry‐aware foundation upon which more testable and biologically relevant molecular design systems can be built.
- Open Access
- 23 July 2026
- Published by Global Science Publishing Pte. Ltd. in Journal of Integrated Social Sciences and Humanities
- p. 0017
From Traits to Training: Personality-Informed Education for Health Professions Students
Health professions education is asked to do several things at once: lay down biomedical knowledge, train clinical skill, and grow communication, teamwork, professionalism, and a capacity for working when the answers are not yet clear. And yet students who sit through the very same curriculum come out looking remarkably different—in how they organise their studying, how they take feedback, how they deal with patients and colleagues, and how they hold up under pressure. Personality traits offer one lens for making sense of that spread, even if their bearing on education is often overstated. This narrative review pulls together the evidence on personality and the main outcomes of students in medicine, nursing, dentistry, pharmacy, rehabilitation, and the other health professions. The sharpest signal in the literature is that conscientiousness follows academic achievement, largely through planning, persistence, the regulation of effort, and simply seeing tasks through to the end. Ties to clinical competence are looser, turning more on the format of the assessment, the stage of training, and the environment in which learning happens. Agreeableness and openness bear on particular corners of empathy and interprofessional learning, whereas neuroticism is the trait most dependably linked to vulnerability under stress, personal distress, and burnout. When it comes to specialty choice, the findings are too tangled to justify prediction at the level of any single individual. Personality assessment can hold modest formative value where it nudges students to reflect on things they are actually in a position to change—self-regulation, communication, engagement with feedback, coping, and reaching out for help—but it has no place standing alone as grounds for admission, ranking, progression, remediation, or the assignment of a career. Handled with care, personality-informed education holds every student to the same competency standards while varying the paths, and the support, that carry them there.
- 23 July 2026
- Published by Global Science Publishing Pte. Ltd. in Journal of Integrated Social Sciences and Humanities
- p. 0017
Open AccessFrom Traits to Training: Personality-Informed Education for Health Professions Students
Health professions education is asked to do several things at once: lay down biomedical knowledge, train clinical skill, and grow communication, teamwork, professionalism, and a capacity for working when the answers are not yet clear. And yet students who sit through the very same curriculum come out looking remarkably different—in how they organise their studying, how they take feedback, how they deal with patients and colleagues, and how they hold up under pressure. Personality traits offer one lens for making sense of that spread, even if their bearing on education is often overstated. This narrative review pulls together the evidence on personality and the main outcomes of students in medicine, nursing, dentistry, pharmacy, rehabilitation, and the other health professions. The sharpest signal in the literature is that conscientiousness follows academic achievement, largely through planning, persistence, the regulation of effort, and simply seeing tasks through to the end. Ties to clinical competence are looser, turning more on the format of the assessment, the stage of training, and the environment in which learning happens. Agreeableness and openness bear on particular corners of empathy and interprofessional learning, whereas neuroticism is the trait most dependably linked to vulnerability under stress, personal distress, and burnout. When it comes to specialty choice, the findings are too tangled to justify prediction at the level of any single individual. Personality assessment can hold modest formative value where it nudges students to reflect on things they are actually in a position to change—self-regulation, communication, engagement with feedback, coping, and reaching out for help—but it has no place standing alone as grounds for admission, ranking, progression, remediation, or the assignment of a career. Handled with care, personality-informed education holds every student to the same competency standards while varying the paths, and the support, that carry them there.
- Open Access
- 4 February 2026
- Published by Frontiers Media SA in Frontiers in Public Health
- Vol. 14
The relationship between medical students’ attitudes toward artificial intelligence and their personality traits: a multicenter study in China
Background: Artificial intelligence (AI) is reshaping healthcare from clinical care and diagnostics to operations and public health and has been accelerated by recent advances in large language models (LLMs), such as ChatGPT. Because healthcare directly concerns human life and well-being, the use of AI must be approached with care. As AI systems are ultimately designed, implemented, and trusted by people, identifying the factors that shape medical students’ attitudes toward AI is critical for safe and effective integration. Methods: In this study, attitudes toward AI were measured with the General Attitudes toward AI Scale (GAAIS), and personality traits were measured with the Big Five Inventory-2 (BFI-2). Along with demographic information, all these data were collected via an online platform from five-year clinical medicine students in Beijing, Shanghai, and Chengdu. Correlation and linear regression were conducted. Results: Analyses indicated that openness and agreeableness were associated with higher scores of positive attitudes toward AI, whereas conscientiousness was associated with lower scores of positive attitudes toward AI. For negative attitudes toward AI, openness and agreeableness were associated with its higher scores, whereas neuroticism was associated with its lower scores. Conclusion: These results suggest that students with greater openness and agreeableness not only view AI more favorably but are also more tolerant of its limitations. In contrast, those high in conscientiousness report fewer positive views, and those high in neuroticism are less tolerant of AI shortcomings. To our knowledge, this is the first study to examine the associations between personality traits and attitudes toward AI in medical students in China, highlighting the need for targeted educational interventions that reflect diverse personality profiles.
- 4 February 2026
- Published by Frontiers Media SA in Frontiers in Public Health
- Vol. 14
Open AccessThe relationship between medical students’ attitudes toward artificial intelligence and their personality traits: a multicenter study in China
Background: Artificial intelligence (AI) is reshaping healthcare from clinical care and diagnostics to operations and public health and has been accelerated by recent advances in large language models (LLMs), such as ChatGPT. Because healthcare directly concerns human life and well-being, the use of AI must be approached with care. As AI systems are ultimately designed, implemented, and trusted by people, identifying the factors that shape medical students’ attitudes toward AI is critical for safe and effective integration. Methods: In this study, attitudes toward AI were measured with the General Attitudes toward AI Scale (GAAIS), and personality traits were measured with the Big Five Inventory-2 (BFI-2). Along with demographic information, all these data were collected via an online platform from five-year clinical medicine students in Beijing, Shanghai, and Chengdu. Correlation and linear regression were conducted. Results: Analyses indicated that openness and agreeableness were associated with higher scores of positive attitudes toward AI, whereas conscientiousness was associated with lower scores of positive attitudes toward AI. For negative attitudes toward AI, openness and agreeableness were associated with its higher scores, whereas neuroticism was associated with its lower scores. Conclusion: These results suggest that students with greater openness and agreeableness not only view AI more favorably but are also more tolerant of its limitations. In contrast, those high in conscientiousness report fewer positive views, and those high in neuroticism are less tolerant of AI shortcomings. To our knowledge, this is the first study to examine the associations between personality traits and attitudes toward AI in medical students in China, highlighting the need for targeted educational interventions that reflect diverse personality profiles.
- Open Access
- 27 November 2025
- Published by Springer Nature in Molecular Biomedicine
- Vol. 6 (1) , 125
T-cell immunoglobulin and mucin domain-containing protein 3–mediated immunomodulation in myeloid cells and keratinocytes in the development of severe acne
Acne vulgaris is a common chronic inflammatory skin disorder with significant clinical and societal impacts. Severe acne (SA), in particular, causes scarring, disfigurement, and psychosocial distress. The underlying pathogenesis of SA remains poorly understood, hindering the development of effective treatments. In our study, single-cell RNA sequencing (scRNA-seq) of samples from SA, acne, and normal skin was performed, 12 clusters were identified from the total cell population, and macrophages were considered as important cell clusters since all gene modules showed high gene activity for macrophages in the scRNA weighted correlation network analysis (WGCNA). By integrating macrophage-specific differentially expressed genes (DEGs) with Mendelian randomization (MR) analysis, it was found that T-cell immunoglobulin and mucin domain-containing protein 3 (TIM3) was identified as a potential immune checkpoint involved in the development of SA. Multiplex immunohistochemistry (mIHC) revealed a decreased tendency for TIM3 + neutrophils, an increased presence of TIM3 + macrophages (especially TIM3 + M2 macrophages), and a reduced population of TIM3 + keratinocytes in SA tissues compared to controls. In human immortalized keratinocytes (HaCaT) cells, TIM3 knockdown led to the upregulation of Propionibacterium acnes (P. acnes) -induced proinflammatory cytokine secretion, and the administration of an anti-Tim3 antibody in P.acnes induced mouse model exacerbated acne-associated inflammation. Collectively, these findings support a role for TIM3 in SA and suggest TIM3 as a potential therapeutic target, while underscoring the immunomodulatory function of keratinocytes and providing directions for future investigations.
- 27 November 2025
- Published by Springer Nature in Molecular Biomedicine
- Vol. 6 (1) , 125
Open AccessT-cell immunoglobulin and mucin domain-containing protein 3–mediated immunomodulation in myeloid cells and keratinocytes in the development of severe acne
Acne vulgaris is a common chronic inflammatory skin disorder with significant clinical and societal impacts. Severe acne (SA), in particular, causes scarring, disfigurement, and psychosocial distress. The underlying pathogenesis of SA remains poorly understood, hindering the development of effective treatments. In our study, single-cell RNA sequencing (scRNA-seq) of samples from SA, acne, and normal skin was performed, 12 clusters were identified from the total cell population, and macrophages were considered as important cell clusters since all gene modules showed high gene activity for macrophages in the scRNA weighted correlation network analysis (WGCNA). By integrating macrophage-specific differentially expressed genes (DEGs) with Mendelian randomization (MR) analysis, it was found that T-cell immunoglobulin and mucin domain-containing protein 3 (TIM3) was identified as a potential immune checkpoint involved in the development of SA. Multiplex immunohistochemistry (mIHC) revealed a decreased tendency for TIM3 + neutrophils, an increased presence of TIM3 + macrophages (especially TIM3 + M2 macrophages), and a reduced population of TIM3 + keratinocytes in SA tissues compared to controls. In human immortalized keratinocytes (HaCaT) cells, TIM3 knockdown led to the upregulation of Propionibacterium acnes (P. acnes) -induced proinflammatory cytokine secretion, and the administration of an anti-Tim3 antibody in P.acnes induced mouse model exacerbated acne-associated inflammation. Collectively, these findings support a role for TIM3 in SA and suggest TIM3 as a potential therapeutic target, while underscoring the immunomodulatory function of keratinocytes and providing directions for future investigations.
- Open Access
Integrated Pan-Cancer Analysis and Experimental Verification of the Roles of Retinoid-Binding Proteins in Breast Cancer
Background: Retinoid-binding proteins (RBPs) regulate retinoid metabolism and signaling, but their roles across human cancers remain incompletely defined. Methods: We conducted a comprehensive analysis using bioinformatics tools and experimental validations, examining RBP expression profiles across cancer types based on data from The Cancer Genome Atlas (TCGA). We employed survival analysis using the Kaplan–Meier method and utilized single-cell RNA sequencing (scRNA-seq) to investigate the roles of RBP4 and RBP7 in the tumor microenvironment. Results: Our analysis revealed significant downregulation of RBPs in multiple cancers, with RBP4 and RBP7 showing notable expression variations linked to tumor stages and grades. Cox analysis identified RBP4 as a protective gene in kidney renal papillary cell carcinoma (KIRP), liver hepatocellular carcinoma (LIHC), and mesothelioma (MESO), while RBP7 exhibited protective effects in breast cancer (BRCA) and uveal melanoma (UVM). Conclusions: This pan-cancer and single-cell integrative analysis highlights the complex roles of RBPs in cancer progression and their potential as prognostic biomarkers, particularly RBP4 and RBP7 in breast cancer. These findings warrant further investigation into the functional mechanisms of RBPs, which may provide valuable strategies for therapeutic interventions.
Open AccessIntegrated Pan-Cancer Analysis and Experimental Verification of the Roles of Retinoid-Binding Proteins in Breast Cancer
Background: Retinoid-binding proteins (RBPs) regulate retinoid metabolism and signaling, but their roles across human cancers remain incompletely defined. Methods: We conducted a comprehensive analysis using bioinformatics tools and experimental validations, examining RBP expression profiles across cancer types based on data from The Cancer Genome Atlas (TCGA). We employed survival analysis using the Kaplan–Meier method and utilized single-cell RNA sequencing (scRNA-seq) to investigate the roles of RBP4 and RBP7 in the tumor microenvironment. Results: Our analysis revealed significant downregulation of RBPs in multiple cancers, with RBP4 and RBP7 showing notable expression variations linked to tumor stages and grades. Cox analysis identified RBP4 as a protective gene in kidney renal papillary cell carcinoma (KIRP), liver hepatocellular carcinoma (LIHC), and mesothelioma (MESO), while RBP7 exhibited protective effects in breast cancer (BRCA) and uveal melanoma (UVM). Conclusions: This pan-cancer and single-cell integrative analysis highlights the complex roles of RBPs in cancer progression and their potential as prognostic biomarkers, particularly RBP4 and RBP7 in breast cancer. These findings warrant further investigation into the functional mechanisms of RBPs, which may provide valuable strategies for therapeutic interventions.
- 24 January 2025
- Published by American Chemical Society (ACS) in Journal of Medicinal Chemistry
- Vol. 68 (6) , 6127-6148
Discovery of First-in-Class FXR and HSD17B13 Dual Modulator for the Treatment of Metabolic Dysfunction-Associated Fatty Liver Disease
Metabolic dysfunction-associated steatohepatitis (MASH) is a complex disease driven by diverse metabolic and inflammatory pathways. Farnesoid X receptor (FXR) is a promising target for MASH due to its role in bile acid and lipid metabolism, while HSD17B13 regulates liver lipid droplet homeostasis. However, the existing HSD17B13 inhibitors have several druglike property challenges due to the common phenolic structure, a key pharmacophore for the HSD17B13 inhibitor. In this study, a two-round high-throughput screening was performed to identify the FXR agonist 2 as the nonphenolic HSD17B13 inhibitor. The multiparameter structural optimization led to the discovery of dual FXR/HSD17B13 modulator 6, with high target selectivity, target tissue distribution, suitable pharmacokinetic properties, and safety profiles. Moreover, even at the lower dose, compound 6 exerted a better therapeutic effect than obeticholic acid (OCA) in multiple MASH models. With attractive pharmacological activity and safety profiles, the dual FXR/HSD17B13 modulator 6 is worthy of further evaluation as a novel anti-MASH agent.
- 24 January 2025
- Published by American Chemical Society (ACS) in Journal of Medicinal Chemistry
- Vol. 68 (6) , 6127-6148
Discovery of First-in-Class FXR and HSD17B13 Dual Modulator for the Treatment of Metabolic Dysfunction-Associated Fatty Liver Disease
Metabolic dysfunction-associated steatohepatitis (MASH) is a complex disease driven by diverse metabolic and inflammatory pathways. Farnesoid X receptor (FXR) is a promising target for MASH due to its role in bile acid and lipid metabolism, while HSD17B13 regulates liver lipid droplet homeostasis. However, the existing HSD17B13 inhibitors have several druglike property challenges due to the common phenolic structure, a key pharmacophore for the HSD17B13 inhibitor. In this study, a two-round high-throughput screening was performed to identify the FXR agonist 2 as the nonphenolic HSD17B13 inhibitor. The multiparameter structural optimization led to the discovery of dual FXR/HSD17B13 modulator 6, with high target selectivity, target tissue distribution, suitable pharmacokinetic properties, and safety profiles. Moreover, even at the lower dose, compound 6 exerted a better therapeutic effect than obeticholic acid (OCA) in multiple MASH models. With attractive pharmacological activity and safety profiles, the dual FXR/HSD17B13 modulator 6 is worthy of further evaluation as a novel anti-MASH agent.
- Open Access
- 5 January 2021
- Published by Springer Nature in Molecular Cancer
- Vol. 20 (1) , 1-7
Bone marrow infiltrated natural killer cells predicted the anti-leukemia activity of MCL1 or BCL2 inhibitors in acute myeloid leukemia
Acute myeloid leukemia (AML) is still incurable due to its heterogeneity and complexity of tumor microenvironment. It is imperative therefore to understand the molecular pathogenesis of AML and identify leukemia-associated biomarkers to formulate effective treatment strategies. Here, we systematically analyzed the clinical characters and natural killer (NK) cells portion in seventy newly-diagnosis (ND) AML patients. We found that the proportion of NK cells in the bone marrow of ND-AML patients could predict the prognosis of patients by analyzing the types and expression abundance of NK related ligands in tumor cells. Furthermore, MCL1 inhibitor but not BCL2 inhibitor combined with NK cell-based immunotherapy could effectively improve the therapeutic efficiency via inhibiting proliferation and inducing apoptosis of AML primary cells as well as cell lines in vitro. There results provide valuable insights that could help for exploring new therapeutic strategies for leukemia treatment.
- 5 January 2021
- Published by Springer Nature in Molecular Cancer
- Vol. 20 (1) , 1-7
Open AccessBone marrow infiltrated natural killer cells predicted the anti-leukemia activity of MCL1 or BCL2 inhibitors in acute myeloid leukemia
Acute myeloid leukemia (AML) is still incurable due to its heterogeneity and complexity of tumor microenvironment. It is imperative therefore to understand the molecular pathogenesis of AML and identify leukemia-associated biomarkers to formulate effective treatment strategies. Here, we systematically analyzed the clinical characters and natural killer (NK) cells portion in seventy newly-diagnosis (ND) AML patients. We found that the proportion of NK cells in the bone marrow of ND-AML patients could predict the prognosis of patients by analyzing the types and expression abundance of NK related ligands in tumor cells. Furthermore, MCL1 inhibitor but not BCL2 inhibitor combined with NK cell-based immunotherapy could effectively improve the therapeutic efficiency via inhibiting proliferation and inducing apoptosis of AML primary cells as well as cell lines in vitro. There results provide valuable insights that could help for exploring new therapeutic strategies for leukemia treatment.