AIM:To investigate the therapeutic effects of bone marrow mesenchymal stem cell-derived exosomes(BMSC-Exos)on diabetic retinopathy(DR)in rats, with a focus on their ability to modulate oxidative stress and inflammatory responses through the TLR4/NF-κB signaling pathway.METHODS:Streptozotocin(STZ)-induced diabetic Sprague-Dawley(SD)rats were randomly divided into four groups: Group 1 [normal+phosphate-buffered saline(PBS)], Group 2(normal+Exos), Group 3(DR+PBS), and Group 4(DR+Exos). At the 8th week after modeling, BMSC-Exos or PBS were injected intravitreally. Retinal tissues were collected at the 16th week for histological analysis [hematoxylin and eosin(HE)staining], apoptosis detection(TUNEL). Oxidative stress markers [8-OHdG, superoxide dismutase(SOD), and glutathione(GSH)] were assessed, and inflammatory cytokines [enzyme-linked immunosorbent assay(ELISA)for interleukin(IL)-6 and tumor necrosis factor-alpha(TNF-α)], and molecular profiling [quantitative polymerase chain reaction(qPCR)for TLR4, NF-κB, and VEGF] were measured.RESULTS:The BMSC-Exos treatment significantly suppressed the activation of the TLR4/NF-κB pathway in DR rats(P<0.01), which was accompanied by reduced retinal vascular leakage(Evans blue assay, P<0.001), decreased apoptosis(TUNEL, P<0.05), and attenuated oxidative stress(elevated SOD and GSH, reduced 8-OHdG, P<0.05). The levels of inflammatory cytokines(IL-6 and TNF-α)were markedly decreased(P<0.01).CONCLUSION:BMSC-Exos alleviate DR by suppressing the TLR4/NF-κB axis, thus reducing oxidative damage, inflammation, and microvascular dysfunction. This research offers a novel therapeutic approach for early-stage DR.

Cremastrae Pseudobulbus Pleiones Pseudobulbus is a Chinese medicinal material used for clearing heat， detoxicifying， resolving phlegm， and dissipating nodules. It is derived from Cremastra appendiculata， Pleione bulbocodioides， or Pleione yunnanensis， as recorded in the 2025 edition of the Pharmacopoeia of the People′s Republic of China， with the former known as Maocigu and the latter two known as Bingqiuzi. Current antitumor studies mainly focus on polysaccharides， water-extracted alcohol-precipitated fractions， and purified extracts of C. appendiculata. Structurally characterized glucomannans suppress breast and liver tumor xenografts and exhibit immunomodulatory effects. The water-extracted alcohol-precipitated fractions inhibit tumor angiogenesis in hepatoma-bearing mice. A purified C. appendiculata extract standardized to five phenanthrene and bibenzyl constituents induces mitochondrial apoptosis in hepatocellular carcinoma cells. Reported pharmacological effects involve cell-cycle regulation， apoptosis， epithelial-mesenchymal transition， angiogenesis， autophagy， and the tumor immune microenvironment， with phosphatidylinositol 3-kinase/protein kinase B/mammalian target of rapamycin （PI3K/Akt/mTOR）， Wnt/β-catenin， and Janus kinase 2/signal transducer and activator of transcription 3 （JAK2/STAT3） signaling pathways being the major mechanism links. Quality evaluation of Cremastrae Pseudobulbus Pleiones Pseudobulbus has progressed from single-marker assays to an integrated approach involving botanical authentication， chromatographic fingerprinting， multicomponent analysis， and evaluation of associations between blood-absorbed constituents and antitumor effects. Militarine and dactylorhin A can be used for botanical identification and batch evaluation， and in vivo exposure data from related orchids are concentrated on militarine. Quantitative associations between these constituents and antitumor effects remain to be established. Clinical evidence mainly focuses on formula-based applications and expert experience. Comparative antitumor data in hepatoma models are available for the commercial specifications of Maocigu and Bingqiuzi， whereas parallel evaluations of pharmacological effects， safety windows， and drug interactions among the three official botanical origins remain insufficient. The major limitations are the incomplete standardization of botanical origins and preparations and the lack of quantitative links among quality markers， systemic exposure， antitumor effects， and toxicity endpoints， which constrain reproducible pharmacological evaluation and clinical translation of Cremastrae Pseudobulbus/Pleiones Pseudobulbus.

Objective:To evaluate the impact of donor characteristics on the prognosis of myelodysplastic syndrome (MDS) patients undergoing haplo-identical transplantation (HIDT) .Methods:A retrospective analysis of 203 MDS patients who received HIDT was conducted to evaluate how donor factors influenced transplant outcomes.Results:In MDS patients undergoing haploidentical transplantation, donors over 50 years were associated with higher EBV reactivation (2-year cumulative incidence 42.9% vs 22.0% for <50 years old; P=0.010). Female donors were linked to increased severe chronic GVHD compared with male donors (2-year incidence 11.9% vs 4.0% ; P=0.017). Additionally, 2-year overall survival (OS) was slightly lower with female donors than male donors (56.6% vs 69.7% ), but the difference was not statistically significant ( P=0.073). Donor-recipient blood type did not affect post-transplant OS or cumulative relapse rates. Donor-recipient kinship analysis revealed that child donors, compared to haploidentical sibling or parent donors, had lower rates of grade Ⅱ–Ⅳ acute GVHD (27.2% vs 45.7% vs 53.5%, P=0.007) and 2-year EBV reactivation (13.9% vs 29.3% vs 38.9%, P=0.001). For donors under 20 years, donor gender did not significantly affect 2-year OS ( P=0.913), relapse-free survival ( P=0.716), or 100-day incidence of grade Ⅱ–Ⅳ acute GVHD ( P=0.359) . Conclusion:For MDS patients undergoing HIDT, donors over 50 should be avoided. Male and child donors are preferred, while donor gender does not significantly affect outcomes if the donor is under 20 years old.

Approximately 20% to 30% of the global workforce is engaged in shift work. As a significant cause of circadian disruption, shift work is closely associated with an increased risk for periodontitis. Nevertheless, how shift work-related circadian disruption functions in periodontitis remains unknown. Herein, we employed a simulated shift work model constructed by controlling the environmental light-dark cycles and revealed that shift work-related circadian disruption exacerbated the progression of experimental periodontitis. RNA sequencing and in vitro experiments indicated that downregulation of the core circadian protein brain and muscle ARNT-like protein 1 (BMAL1) and activation of the Gasdermin D (GSDMD)-mediated pyroptosis were involved in the pathogenesis of that. Mechanically, BMAL1 regulated GSDMD-mediated pyroptosis by suppressing NOD-like receptor protein 3 (NLRP3) inflammasome signaling through modulating nuclear receptor subfamily 1 group D member 1 (NR1D1), and inhibiting Gsdmd transcription via directly binding to the E-box elements in its promoter. GSDMD-mediated pyroptosis accelerated periodontitis progression, whereas downregulated BMAL1 under circadian disruption further aggravated periodontal destruction by increasing GSDMD activity. And restoring the level of BMAL1 by circadian recovery and SR8278 injection alleviated simulated shift work-exacerbated periodontitis via lessening GSDMD-mediated pyroptosis. These findings provide new evidence and potential interventional targets for circadian disruption-accelerated periodontitis.
Pyroptosis/physiology* ; ARNTL Transcription Factors/metabolism* ; Animals ; Periodontitis/etiology* ; Mice ; Phosphate-Binding Proteins/metabolism* ; Shift Work Schedule/adverse effects* ; Intracellular Signaling Peptides and Proteins/metabolism* ; Mice, Inbred C57BL ; Male ; Disease Models, Animal ; Gasdermins

Drug development remains a critical issue in the field of biomedicine. With the rapid advancement of information technologies such as artificial intelligence (AI) and the advent of the big data era, AI-assisted drug development has become a new trend, particularly in predicting drug-target associations. To address the challenge of drug-target prediction, AI-driven models have emerged as powerful tools, offering innovative solutions by effectively extracting features from complex biological data, accurately modeling molecular interactions, and precisely predicting potential drug-target outcomes. Traditional machine learning (ML), network-based, and advanced deep learning architectures such as convolutional neural networks (CNNs), graph convolutional networks (GCNs), and transformers play a pivotal role. This review systematically compiles and evaluates AI algorithms for drug- and drug combination-target predictions, highlighting their theoretical frameworks, strengths, and limitations. CNNs effectively identify spatial patterns and molecular features critical for drug-target interactions. GCNs provide deep insights into molecular interactions via relational data, whereas transformers increase prediction accuracy by capturing complex dependencies within biological sequences. Network-based models offer a systematic perspective by integrating diverse data sources, and traditional ML efficiently handles large datasets to improve overall predictive accuracy. Collectively, these AI-driven methods are transforming drug-target predictions and advancing the development of personalized therapy. This review summarizes the application of AI in drug development, particularly in drug-target prediction, and offers recommendations on models and algorithms for researchers engaged in biomedical research. It also provides typical cases to better illustrate how AI can further accelerate development in the fields of biomedicine and drug discovery.

Objective : To investigate the factors affecting the outcome of high-risk human papillomavirus ( HR- HPV) infection in patients with normal cervix examined by colposcopy in Hefei area and the relationship between persistent HR-HPV infection and cervical cytology． Methods : Data of colposcopy patients were collected from 478 HR-HPV infected patients with normal cervix through colposcopy．Their age，number of sexual partners，contracep- tive methods and other relevant basic information were recorded．Vaginal interferon use，HR-HPV infection at year 1 and year 2，and cervical liquid-based cytology test ( LCT) results were tracked，univariate and multivariate ana- lyses were performed based on basic information，and ROC curves were plotted． Results : The HR-HPV clearance rate at 1 year was 59. 41% ，and the clearance rate at 2 years was 66. 75%．The other 12 types of infection ( 31， 33，35，39，45，51，52，56，58，59，66，68) were more common than the 16 and 18 types．Univariate and mult- ivariate analyses showed that age＞50 years，number of sexual partners ≥2，and history of cervical conectomy in- creased the risk of persistent HR-HPV infection ( χge = 21. 676，P ＜0. 001; χumber of sexual partners = 8. 262，P =0. 004; χistory of cervical conectomy = 11. 267，P = 0. 001 ) ． The risk of HR-HPV infection was significantly lower when condom or vaginal interferon was used ( χondom use = 10. 885，P = 0. 001; χnterferon use = 4. 099，P = 0. 043) ．The area under the ROC curve (AUC) of combined diagnosis of HR-HPV persistent infection was higher than that of single diagnosis，and the AUC of combined diagnosis was 0. 737．Persistent HR-HPV infection was an independent risk factor for abnormal LCT，and the AUC predicted by the model was 0. 755．No cancer was found in patients with persistent HR-HPV infection for 2 years，and the proportion of abnormal LCT was higher than that in patients with negative HR-HPV．The difference was statistically significant ( χ2 = 39. 64，P＜0. 001) ． Conclusion The combined ROC model constructed for patients＞50 years old，with multiple sexual partners，history of cervical surgery， no vaginal interferon use，and no condom use has certain value in predicting persistent HR-HPV infection，and per- sistent HR-HPV infection has predictive value in predicting LCT abnormalities．

Drug development remains a critical issue in the field of biomedicine.With the rapid advancement of information technologies such as artificial intelligence(AI)and the advent of the big data era,AI-assisted drug development has become a new trend,particularly in predicting drug-target associations.To address the challenge of drug-target prediction,AI-driven models have emerged as powerful tools,of-fering innovative solutions by effectively extracting features from complex biological data,accurately modeling molecular interactions,and precisely predicting potential drug-target outcomes.Traditional machine learning(ML),network-based,and advanced deep learning architectures such as convolutional neural networks(CNNs),graph convolutional networks(GCNs),and transformers play a pivotal role.This review systematically compiles and evaluates AI algorithms for drug-and drug combination-target predictions,highlighting their theoretical frameworks,strengths,and limitations.CNNs effectively identify spatial patterns and molecular features critical for drug-target interactions.GCNs provide deep insights into molecular interactions via relational data,whereas transformers increase prediction accu-racy by capturing complex dependencies within biological sequences.Network-based models offer a systematic perspective by integrating diverse data sources,and traditional ML efficiently handles large datasets to improve overall predictive accuracy.Collectively,these AI-driven methods are transforming drug-target predictions and advancing the development of personalized therapy.This review summa-rizes the application of AI in drug development,particularly in drug-target prediction,and offers rec-ommendations on models and algorithms for researchers engaged in biomedical research.It also provides typical cases to better illustrate how AI can further accelerate development in the fields of biomedicine and drug discovery.

Traditional Chinese medicine (TCM) is characterized by complex, multicomponent herbal formulations that challenge the conventional“one drug, one target” paradigm. Network pharmacology, through the construction of multilayered drug-target-disease networks, provides a systematic framework for unraveling TCM’s multitarget and multipathway mechanisms. Recent advancements in artificial intelligence, particularly large language models (LLMs), further enhance data integration, target identification, and clinical decision-making. This review synthesizes current progress in the application of network pharmacology and LLMs in TCM, highlighting their potential to deepen mechanistic insights and optimize drug discovery. By bridging traditional medical wisdom with modern computational tools, this integrative approach aims to advance the scientific validation of TCM and foster innovative healthcare solutions.