中南大学学报(医学版)最新文献

Langerhans cell histiocytosis (LCH) is a group of myeloid neoplastic disorders characterized by infiltration of Langerhans cells, which can accumulate in tissues and cause multisystem manifestations. Erdheim-Chester disease (ECD) is a rare non-LCH histiocytosis, characterized by multisystem infiltration of lipid-laden foamy macrophages. Both ECD and LCH can lead to severe systemic disease, but reports of their overlap remain rare. We describe a female patient with ECD-LCH overlap syndrome presenting predominantly with pericardial effusion. She was admitted to the Department of Respiratory and Critical Care Medicine, Xiangya Hospital, Central South University, on November 10, 2020, due to dyspnea and bilateral lower-extremity edema, with a disease course of 13 months. Computed tomography revealed multi-system involvement, and genetic testing identified a BRAF^V600E mutation. Immunohistochemical staining analysis eventually confirmed ECD-LCH overlap syndrome. Symptomatic treatment was initiated, and follow-up showed stable clinical symptoms. To our knowledge, this is the first reported case worldwide of adult ECD-LCH overlap syndrome with pericardial involvement.

Objectives: Non-suicidal self-injury (NSSI) among college students has become a significant public health concern, highlighting the need for effective early identification tools. This study aims to construct a predictive model for NSSI among college students using the least absolute shrinkage and selection operator (LASSO) regression analysis.

Methods: From April to June 2022, an online questionnaire survey was conducted among college students in 6 provinces: Hunan, Jiangxi, Hubei, Shandong, Guangdong, and Jilin. Sociodemographic information was collected, along with assessments using the Adolescent Non-suicidal Self-injury Assessment Questionnaire, Patient Health Questionnaire-9, Anger Rumination Scale, Multiple Forms of Violence Scale, Childhood Trauma Questionnaire-28 item Short Form, and Community Assessment of Psychic Experiences. LASSO regression analysis was performed to identify predictors of NSSI, construct the predictive model, and develop a nomogram. Calibration curves and receiver operating characteristic (ROC) curves were used to evaluate the calibration and discrimination of the model.

Results: A total of 4 121 college students participated in this study, among whom 650 reported NSSI behaviors, yielding a detection rate of 15.8%. LASSO regression identified 5 predictors of NSSI: Experiences of bullying in primary school, history of alcohol use, depressive symptoms, anger rumination, and psychotic-like experiences. The predictive model was expressed as: Risk of NSSI = (bullying in primary school × 0.41) + (history of alcohol use × 0.76) + (depressive symptoms × 0.08) + (anger rumination × 0.04) + (psychotic-like experiences × 0.05). The area under the curve (AUC) of the predictive model was 0.782 for the training set and 0.769 for the testing set. Calibration curves indicated good agreement between predicted and observed values.

Conclusions: The predictive model demonstrated strong predictive ability and was visualized using a nomogram. This model can be used to assess the risk of NSSI among college students based on identified risk factors and may assist clinicians and educators in identifying high-risk individuals for early interventions.

Breast cancer is one of the most common and fatal malignancies among women worldwide, and its treatment efficacy is often limited by drug resistance and the presence of undruggable targets. Traditional small-molecule drugs have difficulty effectively modulating certain critical targets such as transcription factors and non-coding RNAs, necessitating new therapeutic strategies. Proteolysis-targeting chimeras (PROTACs) function by recruiting pathogenic proteins to the cellular ubiquitin-proteasome system, thereby inducing their specific degradation. In contrast, ribonuclease-targeting chimeras (RIBOTACs) utilize small-molecule ligands but bind to RNA and direct endogenous RNases to selectively degrade pathogenic RNA molecules. By employing a "degradation rather than inhibition" mechanism, targeting chimera technology broadens the druggable landscape and offers a novel precision therapeutic strategy for breast cancer, particularly for refractory and drug-resistant cases. This approach not only overcomes the limitations of traditional drugs, such as the absence of suitable binding sites or poor selectivity, but also reduces required dosages and potential adverse effects. Recent studies have preliminarily demonstrated the therapeutic potential of PROTACs and RIBOTACs in breast cancer, encompassing target design, mechanistic investigation, and preclinical as well as early clinical applications. Research into these technologies reveals their ability to tackle previously undruggable targets, thereby providing theoretical support for the development of safer and more effective precision therapies for breast cancer. In the future, with advances in drug delivery systems and clinical trials, PROTACs and RIBOTACs are expected to be used synergistically with immunotherapy and chemotherapy, offering breast cancer patients more promising comprehensive treatment options and potentially driving oncology toward broader intervention of undruggable targets.

Objectives: Keloids are fibrotic skin disorders characterized by excessive collagen deposition and a high recurrence rate, closely associated with inflammatory mediators. However, existing epidemiological studies are limited by confounding factors and reverse causality, making it difficult to establish causation. This study aims to investigate the causal relationship between circulating cytokines and keloids using Mendelian randomization analysis.

Methods: Significant single nucleotide polymorphisms (SNPs) associated with circulating cytokines (exposures) and keloids (outcomes) were extracted from genome-wide association study (GWAS) summary datasets. Eligible SNPs were selected as instrumental variables (IVs). Exposure data were derived from a cytokine GWAS including 8 293 Finnish participants, and outcome data from a keloid GWAS based on the UK Biobank. The inverse-variance weighted (IVW) method served as the primary analytical approach to estimate causal effects, supplemented by weighted median (WME), MR-Egger regression, and other sensitivity analyses. Horizontal pleiotropy was assessed using MR-Egger regression and the MR pleiotropy residual sum and outlier (MR-PRESSO) test, while Cochran's Q test evaluated heterogeneity. Leave-one-out analysis was used to verify robustness and consistency. A reverse MR analysis was also conducted, with keloid as the exposure and cytokines as outcomes, to rule out reverse causation.

Results: IVW analysis identified significant positive causal associations between two cytokines and keloids-macrophage migration inhibitory factor (MIF) [odds ratio (OR)=2.081, 95% confidence interval (CI) 1.219 to 3.552, P=0.007] and monocyte chemoattractant protein-1 (MCP-1) (OR=1.673, 95% CI 1.036 to 2.701, P=0.035). Conversely, stem cell factor (SCF) showed a negative causal relationship with keloids (OR=0.518, 95% CI 0.269 to 0.998, P=0.049). Results from the MR-Egger and weighted median analyses were consistent with IVW findings. No evidence of horizontal pleiotropy was observed (P>0.05). Except for interleukin-6 (P=0.014), no heterogeneity was detected in other cytokines. Leave-one-out analysis further confirmed the robustness of the causal associations. In reverse MR analysis, keloids were causally related only to β-nerve growth factor (beta-NGF) (OR=1.048, 95% CI 1.002 to 1.095, P=0.039), with no heterogeneity or pleiotropy detected in most cytokines (P>0.05).

Conclusions: MIF and MCP-1 exhibit positive causal associations with keloid formation, while SCF shows a negative causal relationship. These findings provide new evidence for the causal involvement of inflammatory cytokines in keloid pathogenesis and offer potential molecular targets for developing novel keloid therapies.

Objectives: Accurate identification of risk factors for precocious puberty is essential for clinical diagnosis and management, yet the performance of natural language processing methods applied to unstructured electronic medical record (EMR) data remains to be fully evaluated. This study aims to assess the performance of a prompt engineering method for extracting individual risk factors of precocious puberty from EMRs.

Methods: Based on the capacity and role-insight-statement-personality-experiment (CRISPE) prompt framework, both simple and optimized prompts were designed to guide the large language model GLM-4-9B in extracting 10 types of risk factors for precocious puberty from 653 EMRs. Accuracy, precision, recall, and F1-score were used as evaluation metrics for the information extraction task.

Results: Under simple and optimized prompt conditions, the overall accuracy, precision, recall, and F1-score of the model were 84.18%, 98.09%, 81.99%, and 89.32% versus 97.15%, 98.31%, 98.16%, and 98.23%, respectively. The optimized prompts achieved more stable performance across age (<9 years vs ≥9 years) and visit-time (<2023 vs ≥2023) subgroups compared with simple prompts. The accuracy range for extracting each risk factor was 60.03%-97.24%, while with optimized prompts, the range improved to 92.19%-99.85%. The largest performance improvement occurred for "beverage intake" (60.03% vs 92.19%), and the smallest for "maternal age of menarche" (97.24% vs 99.23%). In comparing distributions among simple prompts, optimized prompts, and ground truth, statistically significant differences were observed for snack intake, beverage intake, soy milk intake, honey intake, supplement use, tonic use, sleep quality, and sleeping with the light on (all P<0.001), while exercise (P=0.966) and maternal menarche age (P=0.952) showed no significant differences.

Conclusions: Compared with simple prompts, optimized prompts substantially improved the extraction performance of individual risk factors for precocious puberty from EMRs, underscoring the critical role of prompt engineering in enhancing large language model performance.