Xilin Shen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xilin Shen has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Kexin Chen
Department of Epidemiology and Biostatistics, Key Laboratory of Molecular Cancer Epidemiology, National Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin, China
2 shared publications - Yan Sun
Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Tianjin, China
2 shared publications - Huizhi Li
Department of Pathology National Clinical Research Center for Cancer Tianjin's Clinical Research Center for Cancer Tianjin Key Laboratory of Digestive Cancer Tianjin Medical University Cancer Institute & Hospital Tianjin China
1 shared publication - Jia Deng
Department of Pathology National Clinical Research Center for Cancer Tianjin's Clinical Research Center for Cancer Tianjin Key Laboratory of Digestive Cancer Tianjin Medical University Cancer Institute & Hospital Tianjin China
1 shared publication Jia Xin Li
Department of Pathology, Tianjin’s Clinical Research Center for Cancer, Tianjin Key Laboratory of Digestive Cancer, National Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute & Hospital , Tianjin 300060
1 shared publication- Jiaxin Li
Department of Pathology National Clinical Research Center for Cancer Tianjin's Clinical Research Center for Cancer Tianjin Key Laboratory of Digestive Cancer Tianjin Medical University Cancer Institute & Hospital Tianjin China
1 shared publication
Latest Publications
- Open Access
- 29 September 2026
- Published by Wiley in Advanced Science
MIIP Inhibits Colorectal Cancer Progression by Modulating Neutrophils Infiltration and Neutrophil Extracellular Trap Formation
Neutrophil extracellular traps (NETs) contribute to colorectal cancer (CRC) progression, but the underlying regulatory mechanism remains unknown. This study explored how migration and invasion inhibitory protein (MIIP) influences CRC progression by regulating tumor‐associated neutrophils (TANs) infiltration and NETs formation. On the basis of bioinformatics analysis and clinical CRC samples, we observed that a high density of neutrophils and a high abundance of NETs were correlated with low MIIP expression in CRC tissue as well as distant metastasis and poor prognosis. In vitro co‐culture assays revealed that conditioned medium from CRC cells with downregulated MIIP expression recruited neutrophils to form NETs and that NETs further enhanced the migratory and invasive abilities of CRC cells. Mechanistically, the results of this study demonstrated that MIIP directly bound to Protein Kinase R, which inhibited the Nuclear Factor kappa‐B pathway and Interleukin ‐ 8 expression and accordingly suppressed TANs infiltration and NETs formation. In addition, preliminary results from an intrasplenic mouse model of CRC liver metastasis suggested the potential therapeutic significance of Deoxyribonuclease I (DNase I, a NET inhibitor), and MIIP augmented the inhibition of CRC metastasis by DNase I. This study revealed that MIIP is a regulator of CRC cell‐neutrophil crosstalk and indicated that NETs are potential therapeutic targets in CRC.
- 29 September 2026
- Published by Wiley in Advanced Science
Open AccessMIIP Inhibits Colorectal Cancer Progression by Modulating Neutrophils Infiltration and Neutrophil Extracellular Trap Formation
Neutrophil extracellular traps (NETs) contribute to colorectal cancer (CRC) progression, but the underlying regulatory mechanism remains unknown. This study explored how migration and invasion inhibitory protein (MIIP) influences CRC progression by regulating tumor‐associated neutrophils (TANs) infiltration and NETs formation. On the basis of bioinformatics analysis and clinical CRC samples, we observed that a high density of neutrophils and a high abundance of NETs were correlated with low MIIP expression in CRC tissue as well as distant metastasis and poor prognosis. In vitro co‐culture assays revealed that conditioned medium from CRC cells with downregulated MIIP expression recruited neutrophils to form NETs and that NETs further enhanced the migratory and invasive abilities of CRC cells. Mechanistically, the results of this study demonstrated that MIIP directly bound to Protein Kinase R, which inhibited the Nuclear Factor kappa‐B pathway and Interleukin ‐ 8 expression and accordingly suppressed TANs infiltration and NETs formation. In addition, preliminary results from an intrasplenic mouse model of CRC liver metastasis suggested the potential therapeutic significance of Deoxyribonuclease I (DNase I, a NET inhibitor), and MIIP augmented the inhibition of CRC metastasis by DNase I. This study revealed that MIIP is a regulator of CRC cell‐neutrophil crosstalk and indicated that NETs are potential therapeutic targets in CRC.
- Open Access
- 9 June 2026
- Published by Elsevier in SSRN Electronic Journal
Guideline-grounded Agentic AI Enables Reliable Report-level Diagnostic Reasoning in Oncologic Pathology
Background: Accurate oncologic pathology diagnosis remains cognitively demanding, requiring synthesis of heterogeneous evidence within complex guidelines. Although artificial intelligence has shown promise in pathology, most existing systems are image-centric, whereas large language model (LLM)-based approaches for report-level diagnosis remain limited by hallucination, weak guideline grounding, and insufficient entity-level reasoning. Methods: We developed SAGE-Path (Self-reflective Agentic Guideline-grounded Engine for text-based pathology diagnostic assistance), a framework that integrates retrieval-augmented generation with iterative self-reflective reasoning. The system was grounded in a WHO-derived pathology knowledge base comprising 1,811 tumour entities organized into 48,396 retrievable text units. SAGE-Path was evaluated across multiple pathology tasks, including TNM staging, tumour regression grading, rare-tumour diagnosis, metastatic primary-site inference, and complex diagnostic reasoning. Performance was compared with LLM-only baselines and pathologists. Assisted-reader studies were conducted to assess clinical utility. Findings: SAGE-Path achieved 90.3% accuracy in TNM staging and 99.1% accuracy in tumour regression grading, surpassing junior and senior pathologists in these tasks. In rare-tumour diagnosis and metastatic primary-site inference, SAGE-Path improved diagnostic accuracy by approximately 5–6 percentage points compared with LLM-only baselines and reduced diagnostic errors by about 40%. Mechanistically, retrieval and reflection were complementary: together they eliminated correct-to-error transitions observed with reflection alone, producing evidence-grounded reasoning and actionable recommendations. In assisted-reader evaluations, SAGE-Path improved pathologist performance and reduced completion time. Interpretation: By combining authoritative guideline retrieval with structured self-reflective reasoning, SAGE-Path enables reliable and auditable report-level diagnostic reasoning in oncologic pathology. This framework extends computational pathology beyond image-centric analysis and provides a practical approach for supporting complex diagnostic decision-making in routine pathology practice.
- 9 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessGuideline-grounded Agentic AI Enables Reliable Report-level Diagnostic Reasoning in Oncologic Pathology
Background: Accurate oncologic pathology diagnosis remains cognitively demanding, requiring synthesis of heterogeneous evidence within complex guidelines. Although artificial intelligence has shown promise in pathology, most existing systems are image-centric, whereas large language model (LLM)-based approaches for report-level diagnosis remain limited by hallucination, weak guideline grounding, and insufficient entity-level reasoning. Methods: We developed SAGE-Path (Self-reflective Agentic Guideline-grounded Engine for text-based pathology diagnostic assistance), a framework that integrates retrieval-augmented generation with iterative self-reflective reasoning. The system was grounded in a WHO-derived pathology knowledge base comprising 1,811 tumour entities organized into 48,396 retrievable text units. SAGE-Path was evaluated across multiple pathology tasks, including TNM staging, tumour regression grading, rare-tumour diagnosis, metastatic primary-site inference, and complex diagnostic reasoning. Performance was compared with LLM-only baselines and pathologists. Assisted-reader studies were conducted to assess clinical utility. Findings: SAGE-Path achieved 90.3% accuracy in TNM staging and 99.1% accuracy in tumour regression grading, surpassing junior and senior pathologists in these tasks. In rare-tumour diagnosis and metastatic primary-site inference, SAGE-Path improved diagnostic accuracy by approximately 5–6 percentage points compared with LLM-only baselines and reduced diagnostic errors by about 40%. Mechanistically, retrieval and reflection were complementary: together they eliminated correct-to-error transitions observed with reflection alone, producing evidence-grounded reasoning and actionable recommendations. In assisted-reader evaluations, SAGE-Path improved pathologist performance and reduced completion time. Interpretation: By combining authoritative guideline retrieval with structured self-reflective reasoning, SAGE-Path enables reliable and auditable report-level diagnostic reasoning in oncologic pathology. This framework extends computational pathology beyond image-centric analysis and provides a practical approach for supporting complex diagnostic decision-making in routine pathology practice.
- Open Access
- 19 March 2026
- Published by Oxford University Press (OUP) in Nucleic Acids Research
- Vol. 54 (6)
A foundation model for nucleotide sequences
Foundation models have demonstrated exceptional performance across diverse downstream tasks. However, in genomics and transcriptomics, the integration of nucleotide sequences with their rich annotations remains underexplored, potentially limiting model generalizability across species and biological contexts. Here, we introduce OmniNA (Omni-applicable foundation model for nucleic acid), a self-supervised generative foundation model trained on 91.7 million nucleotide sequences and their associated annotations, totaling 1076.2 billion bases and 197 million words spanning diverse species. Unlike most existing approaches that focus solely on functional genomic element recognition, OmniNA jointly learns from both sequences and annotations, leveraging their complementarity to enhance semantic understanding and representation learning. We demonstrate that OmniNA captures sequence grammar and annotation semantics, facilitating robust transfer across nucleotide-level tasks. OmniNA can be fine-tuned under natural language paradigms and achieves state-of-the-art or competitive performance in 23 benchmarks, including sequence detection and species classification. Its learned representations also help reveal mutation effects on DNA and RNA processing. We release the model publicly as a community resource for genomics and transcriptomics research. OmniNA represents a step toward advancing foundational modeling by integrating annotation-aware learning, offering a powerful tool for genomics and transcriptomics research.
- 19 March 2026
- Published by Oxford University Press (OUP) in Nucleic Acids Research
- Vol. 54 (6)
Open AccessA foundation model for nucleotide sequences
Foundation models have demonstrated exceptional performance across diverse downstream tasks. However, in genomics and transcriptomics, the integration of nucleotide sequences with their rich annotations remains underexplored, potentially limiting model generalizability across species and biological contexts. Here, we introduce OmniNA (Omni-applicable foundation model for nucleic acid), a self-supervised generative foundation model trained on 91.7 million nucleotide sequences and their associated annotations, totaling 1076.2 billion bases and 197 million words spanning diverse species. Unlike most existing approaches that focus solely on functional genomic element recognition, OmniNA jointly learns from both sequences and annotations, leveraging their complementarity to enhance semantic understanding and representation learning. We demonstrate that OmniNA captures sequence grammar and annotation semantics, facilitating robust transfer across nucleotide-level tasks. OmniNA can be fine-tuned under natural language paradigms and achieves state-of-the-art or competitive performance in 23 benchmarks, including sequence detection and species classification. Its learned representations also help reveal mutation effects on DNA and RNA processing. We release the model publicly as a community resource for genomics and transcriptomics research. OmniNA represents a step toward advancing foundational modeling by integrating annotation-aware learning, offering a powerful tool for genomics and transcriptomics research.