Shuo Xu

Dr. Shuo Xu


Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 June 2025 in JMIR Medical Informatics
Verified
Journal Article
Open access
ZYZhong Yao
Zhong Yao
LDLiantan Duan
Liantan Duan
SXShuo Xu
Shuo Xu
LCLingyi Chi
Lingyi Chi
DSDongfang Sheng
Dongfang Sheng
Background: Research on large language models (LLMs) in the medical field has predominantly focused on models trained with English-language corpora, evaluating their performance within English-speaking contexts. The performances of models trained with non–English language corpora and their performance in non-English contexts remain underexplored. Objective: This study aimed to evaluate the performances of LLMs trained on different languages corpora by using the Chinese National Medical Licensing Examination (CNMLE) as a benchmark and constructed analogous questions. Methods: Under different prompt settings, we sequentially posed questions to 7 LLMs: 2 primarily trained on English-language corpora and 5 primarily on Chinese-language corpora. The models’ responses were compared against standard answers to calculate the accuracy rate of each model. Further subgroup analyses were conducted by categorizing the questions based on various criteria. We also collected error sets to explore patterns of mistakes across different models. Results: Under the zero-shot setting, 6 out of 7 models exceeded the passing level, with the highest accuracy rate achieved by the Chinese LLM Baichuan (86.67%), followed by ChatGPT (83.83%). In the constructed questions, all 7 models exceeded the passing threshold, with Baichuan maintaining the highest accuracy rate (87.00%). In few-shot learning, all models exceeded the passing threshold. Baichuan, ChatGLM, and ChatGPT retained the highest accuracy. While Llama showed marked improvement over previous tests, the relative performance rankings of other models stayed similar to previous results. In subgroup analyses, English models demonstrated comparable or superior performance to Chinese models on questions related to ethics and policy. All models except Llama generally had higher accuracy rates for simple questions than for complex ones. The error set of ChatGPT was similar to those of other Chinese models. Multimodel cross-verification outperformed single model, particularly improving accuracy rate on simple questions. The implementation of dual-model and tri-model verification achieved accuracy rates of 94.17% and 96.33% respectively. Conclusions: At the current level, LLMs trained primarily on English corpora and those trained mainly on Chinese corpora perform similarly well in CNMLE, with Chinese models still outperforming. The performance difference between ChatGPT and other Chinese LLMs are not solely due to communication barriers but are more likely influenced by disparities in the training data. By using a method of cross-verification with multiple LLMs, excellent performance can be achieved in medical examinations.

28 June 2023 in British Journal of Cancer
Verified
Journal Article
Open access
PDF file
SZShujun Zhang
Shujun Zhang
PLPeilong Li
Peilong Li
JLJuan Li
Juan Li
JGJie Gao
Jie Gao
QQQiuchen Qi
Qiuchen Qi
CWChuanxin Wang
Chuanxin Wang
Background: The epigenetic mechanisms involved in the progression of pancreatic ductal adenocarcinoma (PDAC) remain largely unexplored. This study aimed to identify key transcription factors (TFs) through multiomics sequencing to investigate the molecular mechanisms of TFs that play critical roles in PDAC. Methods: To characterise the epigenetic landscape of genetically engineered mouse models (GEMMs) of PDAC with or without KRAS and/or TP53 mutations, we employed ATAC-seq, H3K27ac ChIP-seq, and RNA-seq. The effect of Fos-like antigen 2 (FOSL2) on survival was assessed using the Kaplan–Meier method and multivariate Cox regression analysis for PDAC patients. To study the potential targets of FOSL2, we performed Cleavage Under Targets and Tagmentation (CUT&Tag). To explore the functions and underlying mechanisms of FOSL2 in PDAC progression, we employed several assays, including CCK8, transwell migration and invasion, RT-qPCR, Western blotting analysis, IHC, ChIP-qPCR, dual-luciferase reporter, and xenograft models. Results: Our findings indicated that epigenetic changes played a role in immunosuppressed signalling during PDAC progression. Moreover, we identified FOSL2 as a critical regulator that was up-regulated in PDAC and associated with poor prognosis in patients. FOSL2 promoted cell proliferation, migration, and invasion. Importantly, our research revealed that FOSL2 acted as a downstream target of the KRAS/MAPK pathway and recruited regulatory T (Treg) cells by transcriptionally activating C-C motif chemokine ligand 28 (CCL28). This discovery highlighted the role of an immunosuppressed regulatory axis involving KRAS/MAPK-FOSL2-CCL28-Treg cells in the development of PDAC. Conclusion: Our study uncovered that KRAS-driven FOSL2 promoted PDAC progression by transcriptionally activating CCL28, revealing an immunosuppressive role for FOSL2 in PDAC.

7 September 2024 in Neurotherapeutics
Verified
Journal Article
Open access
XYXiaxin Yang
Xiaxin Yang
JZJianhang Zhang
Jianhang Zhang
ZWZhihao Wang
Zhihao Wang
ZYZhong Yao
Zhong Yao
XYXue Yang
Xue Yang
SXShuo Xu
Shuo Xu
Temporal lobe epilepsy (TLE) is the most prevalent type of focal epilepsy in adults. While comprehensive bioinformatics analyses have facilitated the identification of novel biomarkers in animal models, similar efforts are limited for TLE patients. In the current study, a comprehensive analysis using human transcriptomics datasets GSE205661, GSE190451, and GSE186334 was conducted to reveal differentially expressed genes related to mitochondria (Mito-DEGs). Protein-protein interaction (PPI) network and Least Absolute Shrinkage and Selection Operator (LASSO) regression analyses were performed to identify hub genes. Additional GSE127871 and GSE255223 were utilized to establish the association with hippocampal sclerosis (HS) and seizure frequency, respectively. Single-cell RNA analysis, functional investigation, and clinical verification were conducted. Herein, we reported that the Mito-DEGs in human TLE were significantly enriched in metabolic processes. Through PPI and LASSO analysis, HSDL2 was identified as the hub gene, of which diagnostic potential was further confirmed using independent datasets, animal models, and clinical validation. Subsequent single-cell and functional analyses revealed that HSDL2 expression was enriched and upregulated in response to excessive lipid accumulation in astrocytes. Additionally, the diagnostic efficiency of blood HSDL2 was verified in Qilu cohort. Together, our findings highlight the translational potential of HSDL2 as a biomarker and provide a novel therapeutic perspective for human TLE.

11 December 2024 in CNS Neuroscience & Therapeutics
Verified
Journal Article
Open access
XGXing Guo
Xing Guo
SGShuhua Guo
Shuhua Guo
FTFeng Tian
Feng Tian
ZGZijie Gao
Zijie Gao
YFYang Fan
Yang Fan
SXShuo Xu
Shuo Xu
Background: The presence of glioma stem cells (GSCs) and the occurrence of mesenchymal phenotype transition contribute to the miserable prognosis of glioblastoma (GBM). Cellular communication network factor 1 (CCN1) is upregulated within various malignancies and associated with cancer development and progression, while the implications of CCN1 in the phenotype transition and tumorigenicity of GSCs remain unclear.Methods: Data for bioinformatic analysis were obtained from The Cancer Genome Atlas (TCGA) and Chinese Glioma Genome Atlas (CGGA) databases. A range of primary GBM and GSC cell models were then used to demonstrate the regulatory role of CCN1 via the phenotype validation, tumor sphere formation assays, extreme limiting dilution assays (ELDA), and transwell assays. To screen out the downstream signaling pathway, we employed high‐throughput RNA‐seq. Intracranial xenograft GSC mouse models were used to investigate the role of CCN1 in vivo.Results: Among the CCN family members, CCN1 was highly expressed in MES‐GBM/GSCs and was correlated with a poor prognosis. Both in vitro and in vivo assays indicated that knockdown of CCN1 in MES‐GSCs reduced the tumor stemness, proliferation, invasion, and tumorigenicity, whereas CCN1 overexpression in PN‐GSCs exhibited the opposite effects. Mechanistically, CCN1 triggered the FAK/STAT3 signaling in autocrine and paracrine manners to upregulate the expression of S100A8. Knockdown of S100A8 inactivated NF‐κB/p65 pathway and significantly suppressed the tumorigenesis of MES‐GSCs.Conclusion: Our findings reveal that CCN1 may be an important factor in the enhanced invasiveness and MES phenotype transition of GSCs and highlight the potential to target CCN1 for treating GBM.

9 May 2025 in Frontiers in Neurology
Verified
Journal Article
Open access
PDF file
SCSi Chen
Si Chen
XYXiaxin Yang
Xiaxin Yang
SXShujun Xu
Shujun Xu
BLBaomin Li
Baomin Li
CLChao Li
Chao Li
XZXiuhe Zhao
Xiuhe Zhao
Introduction: Vagus nerve stimulation (VNS) has garnered widespread application in patients with drug-resistant epilepsy (DRE), while the efficacy and prognostic factors of VNS in DRE remain elusive. Moreover, clinical determinants associated with rapid response to VNS have never been uncovered. This study aimed to elucidate factors influencing efficacy and rapid response to VNS.Methods: A consecutive series of patients with DRE undergoing VNS surgery from January 2014 to December 2023 was collected to describe VNS efficacy. Both univariate and multivariate analyses were performed to identify statistically significant prognostic factors, and a predictive model was developed. Furthermore, we examined clinical determinants of rapid/slow response to VNS and VNS current changes.Results: A total of 65 patients underwent VNS implantation. Seizure frequency significantly decreased post-VNS, with mean seizure reduction rates of 35.7, 49.0, 48.5, 52.8, 63.2, and 66.8% at 6 (n = 65), 12 (n = 65), 24 (n = 50), 36 (n = 40), 60 (n = 31), and 84 (n = 19) months, respectively. At final follow-up, 61.5% (40/65) were responders (50–100% seizure reduction), and 10.8% (7/65) achieved seizure freedom for ≥1 year. Univariate analysis identified age at seizure onset ≥6 years (p = 0.003), baseline seizure frequency ≤30/month (p = 0.001), focal seizures (p = 0.002), developmental and epileptic encephalopathies (p = 0.037), and surgical history (p < 0.001) as significant prognostic factors. Multivariate analysis confirmed age at seizure onset ≥6 years (OR: 5.726, p = 0.039), baseline seizure frequency ≤30/month (OR: 4.697, p = 0.048), and focal seizures (OR: 4.791, p = 0.025) as independent predictors, enabling the development of a predictive model for VNS efficacy. Additionally, among responders, the median response duration was 6 months (range: 1–60 months), with baseline seizure frequency ≤30/month significantly associated with rapid response of VNS in DRE (<6 months, p = 0.033).Conclusion: Vagus nerve stimulation is effective for treating DRE, with efficacy increasing with follow-up duration. Age at seizure onset ≥6 years, baseline seizure frequency ≤30/month, and focal seizure were predictive of VNS success, underscoring the need for careful preoperative assessment of patients with DRE before VNS surgery.

16 December 2021 in Frontiers in Oncology
Verified
Journal Article
Open access
XGXing Guo
Xing Guo
TCTeng Chen
Teng Chen
SCShiming Chen
Shiming Chen
CSChao Song
Chao Song
DSDezhi Shan
Dezhi Shan
SXShuo Xu
Shuo Xu
As an oncogenic somatic variant, telomerase reverse transcriptase promoter (TERTp) mutations are frequently observed in adult glioblastoma (GBM). Alternatively, we report the first case of glioblastoma with TERT amplification accompanied by multiple TERT and FGFR2 gene fusions instead of TERTp mutation. A 55-year-old woman presented with dizziness, headache, and diplopia for three weeks. Magnetic resonance imaging (MRI) demonstrated a heterogeneously enhancing lobulated mass centered in the pineal region. Partial tumor resection and ventriculoperitoneal shunt were achieved, and the residual tumor was then treated with standard radiation. The tumor was diagnosed as GBM, IDH-wild type, WHO grade IV, and the Ki67 proliferation index was high (30–40%). Intriguingly, TERT amplification without TERTp mutation was identified via next generation sequencing (NGS). Further analysis revealed multiple TERT (TERT–NUBPL, MARCH6–TERT, and CJD4–TERT) and FGFR2 (CXCL17–FGFR2, SIPA1L3–FGFR2, FGFR2–SIPA1L3, and FGFR2–CEACAM1) gene fusions. After the surgery, the patient’s condition deteriorated rapidly due to the malignant nature of the tumor and she died with an overall survival of 3 months. Our report provides the molecular clue for a novel telomerase activation and maintenance mechanism in GBM.