Objective: To develop QingNangTCM, a specialized large language model (LLM) tailored for expert-level traditional Chinese medicine (TCM) question-answering and clinical reasoning, addressing the scarcity of domain-specific corpora and specialized alignment. Methods: We constructed QnTCM_Dataset, a corpus of 100 000 entries, by integrating data from ShenNong_TCM_Dataset and SymMap v2.0, and synthesizing additional samples via retrieval-augmented generation (RAG) and persona-driven generation. The dataset comprehensively covers diagnostic inquiries, prescriptions, and herbal knowledge. Utilizing P-Tuning v2, we fine-tuned the GLM-4-9B-Chat backbone to develop QingNangTCM. A multi-dimensional evaluation framework, assessing accuracy, coverage, consistency, safety, professionalism, and fluency, was established using metrics such as bilingual evaluation understudy (BLEU), recall-oriented understudy for gisting evaluation (ROUGE), metric for evaluation of translation with explicit ordering (METEOR), and LLM-as-a-Judge with expert review. Qualitative analysis was conducted across four simulated clinical scenarios: symptom analysis, disease treatment, herb inquiry, and failure cases. Baseline models included GLM-4-9B-Chat, DeepSeek-V2, HuatuoGPT-II (7B), and GLM-4-9B-Chat (freeze-tuning). Results: QingNangTCM achieved the highest scores in BLEU-1/2/3/4 (0.425/0.298/0.137/0.064), ROUGE-1/2 (0.368/0.157), and METEOR (0.218), demonstrating a balanced and superior normalized performance profile of 0.900 across the dimensions of accuracy, coverage, and consistency. Although its ROUGE-L score (0.299) was lower than that of HuatuoGPT-II (7B) (0.351), it significantly outperformed domain-specific models in expert-validated win rates for professionalism (86%) and safety (73%). Qualitative analysis confirmed that the model strictly adheres to the “symptom-syndrome-pathogenesis-treatment” reasoning chain, though occasional misclassifications and hallucinations persisted when dealing with rare medicinal materials and uncommon syndromes. Conclusion Combining domain-specific corpus construction with parameter-efficient prompt tuning enhances the reasoning behavior and domain adaptation of LLMs for TCM-related tasks. This work provides a technical framework for the digital organization and intelligent utilization of TCM knowledge, with potential value for supporting diagnostic reasoning and medical education.

Current epilepsy prediction methods are not effective in characterizing the multi-domain features of complex long-term electroencephalogram (EEG) data, leading to suboptimal prediction performance. Therefore, this paper proposes a novel multi-scale sparse adaptive convolutional network based on multi-head attention mechanism (MS-SACN-MM) model to effectively characterize the multi-domain features. The model first preprocesses the EEG data, constructs multiple convolutional layers to effectively avoid information overload, and uses a multi-layer perceptron and multi-head attention mechanism to focus the network on critical pre-seizure features. Then, it adopts a focal loss training strategy to alleviate class imbalance and enhance the model's robustness. Experimental results show that on the publicly created dataset (CHB-MIT) by MIT and Boston Children's Hospital, the MS-SACN-MM model achieves a maximum accuracy of 0.999 for seizure prediction 10 ~ 15 minutes in advance. This demonstrates good predictive performance and holds significant importance for early intervention and intelligent clinical management of epilepsy patients.
Humans ; Electroencephalography/methods* ; Epilepsy/physiopathology* ; Neural Networks, Computer ; Seizures/physiopathology* ; Signal Processing, Computer-Assisted ; Algorithms

Image 1.

In general, bioassay-guided fractionation and isolation of bioactive constituents from botanical materials frequently ended up with the reward of a single compound. However, botanical materials typically exert their therapeutic actions through multi-pathway effects due to the intrinsic complex nature of chemical constituents. In addition, the content of bioactive compounds in botanical materials is largely dependent on humidity, temperature, soil, especially geographical origins, from which rapid and accurate identification of plant materials is pressingly needed. These long-standing obstacles collectively impede the deep exploitation and application of these versatile natural sources. To address the challenges, a new paradigm integrating biogravimetric analyses and machine learning-driven origin classification (BAMLOC) was developed. The biogravimetric analyses are based on absolute qHNMR quantification and in vivo zebrafish model-assisted activity index calculation, by which bioactive substance groups jointly responsible for the bioactivities in all fractions are pinpointed before any isolation effort. To differentiate origin-different botanical materials varying in the content of bioactive substance groups, principal component analysis, linear discriminant analysis, and hierarchical cluster analysis in conjunction with supervised support vector machine are employed to classify and predict production areas based on the detection of volatile organic compounds by E-nose and GC-MS. Expanding BAMLOC to Codonopsis Radix enables the identification of polyacetylenes and pyrrolidine alkaloids as the bioactive substance group for immune restoration effect and accurately determines the origins of plants. This study advances the toolbox for the discovery of bioactive compounds from complex mixtures and lays a more definitive foundation for the in-depth utilization of botanical materials.

Objective: To examine the causal association and potential mechanisms between bone mineral density in different age groups and primary malignant bone tumor based on two sample Mendelian randomization (MR), so as to provide a reference for the prevention and treatment of primary malignant bone tumor. Methods: The genome-wide association study (GWAS) of bone mineral density was obtained from the GEFOS database,which included 66 628 subjects divided into five age groups (0-15, 15-30, 30-45, 45-60, and >60 years) based on the phases of human bone development. The GWAS of primary malignant bone tumor was sourced from the FinnGen database, including 648 cases and 378 749 controls. Using bone mineral density of five age groups as the exposure and primary malignant bone tumor as the outcome, an MR analysis was performed with the inverse-variance weighted (IVW) method. Sensitivity analysis were conducted using Cochran's Q test, MR-Egger regression, MR-PRESSO test and MR Steiger test. The potential mechanisms underlying the causal association between bone density and primary malignant bone tumors were explored using Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. Results: The MR analysis results showed that there was a negative causal association between bone density and primary malignant bone tumors in the 30-45 age group (OR=0.301, 95%CI: 0.126-0.721). No statistically significant associations between bone density and primary malignant bone tumors were found in the 0-15, 15-30, 45-60, and >60 age groups (all P>0.05). Sensitivity analysis did not detect heterogeneity, pleiotropy (all P>0.05) and reverse causality. KEGG enrichment analysis revealed that genes highly associated with bone density and primary malignant bone tumors were enriched in the mTOR signaling pathway and the Wnt signaling pathway, among which Low Density lipoprotein Receptor Related protein 5 and Wnt Family Member 16 are key regulatory genes. Conclusion The decrease in bone mineral density among individuals aged 30-45 may increase the risk of primary malignant bone tumors through the mTOR signaling pathway and the Wnt signaling pathway.

The aging microenvironment, as a key driver of tumorigenesis and progression, plays a critical role in tumor immune regulation through one of its core features-the senescence-associated secretory phenotype (SASP). SASP consists of a variety of interleukins, chemokines, proteases, and growth factors. It initially induces surrounding cells to enter a state of senescence through paracrine mechanisms, thereby creating a sustained inflammatory stimulus and signal amplification effect within the tissue microenvironment. Furthermore, these secreted factors activate key signaling pathways such as NF-κB, cGAS-STING, and mTOR, which regulate the expression of immune-related molecules (such as PD-L1) and promote the recruitment of immunosuppressive cells, including regulatory T cells and myeloid-derived suppressor cells. This process ultimately contributes to the formation of an immunosuppressive tumor microenvironment. Furthermore, the article explores potential anti-tumor immunotherapy strategies targeting SASP and its associated molecular mechanisms, including approaches to inhibit SASP secretion or eliminate senescent cells. Although these strategies have shown promise in certain tumor models, the high heterogeneity among tumor types may result in varied responses to SASP-targeted therapies. This highlights the need for further research into adaptive stratification and personalized treatment approaches. Targeting immune regulatory mechanisms in the aging microenvironment-particularly SASP-holds great potential for advancing future anti-tumor therapies.

Dysfunction of the interoceptive system is recognized as an important component of clinical symptoms, including anxiety, depression, psychosis, and other mental disorders. Non-invasive neuromodulation is an emerging clinical intervention approach, and over the past decade, research on non-invasive neuromodulation aimed at regulating interoception has rapidly developed. This review first outlines the pathways of interoceptive signals and assessment methods, then summarizes the interoceptive abnormalities in psychiatric disorders and current studies for non-invasive neuromodulation targeting interoception, including intervention modes, target sites, interoceptive measures, and potential neurobiological mechanisms. Finally, we discuss significant research challenges and future directions.
Humans ; Interoception/physiology* ; Mental Disorders/therapy*

Objective: To investigate the causal association between amino acids and the primary malignant bone tumor and its underlying mechanism. Methods: Genome-wide association study (GWAS) data of glycine, serine, arginine, glutamine, methionine, and leucine was sourced from the IEU OpenGWAS database and the GWAS Catalog. GWAS data of primary malignant bone tumor were obtained from the FinnGen database. Using each of the six amino acids as the exposure and primary malignant bone tumor as the outcome, two-sample Mendelian randomization (MR) analysis was performed with the inverse-variance weighted method as the primary approach. Multivariable MR analysis was employed to control for collinearity among amino acids. Sensitivity analyses were conducted using Cochran's Q test, MR-Egger regression and the MR Steiger test. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis and protein-protein interaction network analysis were explored to explore potential mechanisms and identify key genes. Results: MR analysis results indicated a statistically significant causal association between glycine and primary malignant bone tumor (OR=1.719, 95%CI: 1.083-2.728). No significant causal associations were found for the other five amino acids (all P>0.05). Multivariable MR analysis revealed that, after adjusting for the other five amino acids, confirmed a positive causal association between glycine and primary malignant bone tumor (OR=1.512, 95%CI: 1.125-2.031). Sensitivity analyses revealed no significant heterogeneity, horizontal pleiotropy, or reverse causality (all P>0.05). Genes associated with both glycine metabolism and primary malignant bone tumor were enriched in the JAK-STAT signaling pathway, with serine hydroxymethyltransferase 2 (SHMT2) identified as a key gene. Conclusion Higher glycine levels may increase the risk of primary malignant bone tumor via the SHMT2-JAK-STAT pathway.

Objective: To constructCSF1R+/-mice and to analyze their genotypes, so as to provide animal model basis for disease pathological mechanism and drug target. Methods : A linearized targeting vector was designed according to Cre/Loxp system. A Loxp site was inserted upstream of the 5th exon of theCSF1Rgene, and a neomycin resistance box with Loxp sites on both sides was inserted downstream of the 5th exon. The linearized targeting vector was electroporated into embryonic stem cells. The correctly targeted embryonic stem cells were injected into the blastocysts of C57BL/6J mice to obtain chimeric mice, which were bred with Zp3-Cre mice. The newborn mice were numbered 9-14 days after birth and their tails were cut. The DNA of the mice was extracted, and the genotype of the mice was identified by polymerase chain reaction and agarose gel electrophoresis. The expression of CSF1R in mouse macrophages was detected by flow cytometry. The expression of CSF1R in mouse tissues was detected by Western blot. Results: The results of agarose gel electrophoresis showed that 453 bp bands were amplified in wild type mice, and 453 bp and 650 bp bands were amplified in heterozygous mice. The results of flow cytometry showed that the expression of CSF1R in peritoneal macrophages and bone marrow-derived macrophages of CSF1R heterozygous mice was lower than that of WT group(P<0.05). The results of Western blot showed that the expression of CSF1R in spleen, kidney and brain tissue of CSF1R heterozygous group was lower than that of WT group(P<0.05). Conclusion CSF1R+/-mice are successfully constructed, reproduced and identified, which provides an animal model basis for further revealing the potential mechanism of CSF1R in immune regulation.

Objective:To screen the key genetic,diagnostic and therapeutic targets of chronic obstructive pulmonary disease(COPD)patients by using microarray datasets and Mendelian randomization(MR)method,and to provide the evidence for clinical diagnosis and treatment of COPD.Methods:Four COPD gene expression profile datasets were obtained from the Gene Expression Omnibus(GEO)database.The data were processed and normalized using R software,and differentially expressed genes(DEGs)were screened.MR analysis was performed to explore the causal relationship between COPD and expression quantitative trait loci(eQTL),intersection with DEGs was taken to identify potential key targets.Gene Set Enrichment Analysis(GSEA),Gene Ontology(GO)functional enrichment analysis,and Kyoto Encyclopedia of Genes and Genomes(KEGG)signaling pathway enrichment analysis were conducted to investigate the functional roles and pathways of the key targets,external datasets were used to validate their expression.Results:A total of 1 571 DEGs were screened,including 820 upregulated genes and 751 downregulated genes.MR analysis identified 286 COPD-related genes,and intersection with DEGs revealed 3 upregulated genes:diacylglycerol kinase gamma(DGKG),neurofilament heavy polypeptide(NEFH),and Fc receptor like B(FCRLB);and 6 downregulated genes:STEAP4 metalloreductase(STEAP4),pleckstrin homology domain containing family F member 2(PLEKHF2),CD3d molecule(CD3D),transgelin 2(TAGLN2),tripartite motif containing 22(TRIM22),and ribosomal protein L9(RPL9).The biological function analysis results indicated that these genes were mainly involved in pathways such as iron ion transport into the cells,oxidoreductase activity,primary immunodeficiency,and Th1 and Th2 cell differentiation.The MR analysis results confirmed the causal relationship between these targets and COPD.The external validation results showed that compared with healthy controls,the expression level of FCRLB in COPD samples was significantly increased(P<0.01),while the expression levels of CD3D and RPL9 were significantly decreased(P<0.05 or P<0.01),which was consistent with the MR analysis results,highlighting the reliability of this study.Conclusion:DGKG,NEFH,FCRLB,STEAP4,PLEKHF2,CD3D,TAGLN2,TRIM22,and RPL9 may serve as important regulatory factors and clinical diagnostic/therapeutic targets in the pathogenesis of COPD,providing clues for early screening,diagnosis,and targeted treatment of COPD.