In recent years, the application of artificial intelligence (AI) in the field of biology has witnessed remarkable advancements. Among these, the most notable achievements have emerged in the domain of protein structure prediction and design, with AlphaFold and related innovations earning the 2024 Nobel Prize in Chemistry. These breakthroughs have transformed our ability to understand protein folding and molecular interactions, marking a pivotal milestone in computational biology. Looking ahead, it is foreseeable that the accurate prediction of various physicochemical properties of proteins—beyond static structure—will become the next critical frontier in this rapidly evolving field. One of the most important protein properties is thermodynamic stability, which refers to a protein’s ability to maintain its native conformation under physiological or stress conditions. Accurate prediction of protein stability, especially upon single-point mutations, plays a vital role in numerous scientific and industrial domains. These include understanding the molecular basis of disease, rational drug design, development of therapeutic proteins, design of more robust industrial enzymes, and engineering of biosensors. Consequently, the ability to reliably forecast the stability changes caused by mutations has broad and transformative implications across biomedical and biotechnological applications. Historically, protein stability was assessed via experimental methods such as differential scanning calorimetry (DSC) and circular dichroism (CD), which, while precise, are time-consuming and resource-intensive. This prompted the development of computational approaches, including empirical energy functions and physics-based simulations. However, these traditional models often fall short in capturing the complex, high-dimensional nature of protein conformational landscapes and mutational effects. Recent advances in machine learning (ML) have significantly improved predictive performance in this area. Early ML models used handcrafted features derived from sequence and structure, whereas modern deep learning models leverage massive datasets and learn representations directly from data. Deep neural networks (DNNs), graph neural networks (GNNs), and attention-based architectures such as transformers have shown particular promise. GNNs, in particular, excel at modeling spatial and topological relationships in molecular structures, making them well-suited for protein modeling tasks. Furthermore, attention mechanisms enable models to dynamically weigh the contribution of specific residues or regions, capturing long-range interactions and allosteric effects. Nevertheless, several key challenges remain. These include the imbalance and scarcity of high-quality experimental datasets, particularly for rare or functionally significant mutations, which can lead to biased or overfitted models. Additionally, the inherently dynamic nature of proteins—their conformational flexibility and context-dependent behavior—is difficult to encode in static structural representations. Current models often rely on a single structure or average conformation, which may overlook important aspects of stability modulation. Efforts are ongoing to incorporate multi-conformational ensembles, molecular dynamics simulations, and physics-informed learning frameworks into predictive models. This paper presents a comprehensive review of the evolution of protein thermodynamic stability prediction techniques, with emphasis on the recent progress enabled by machine learning. It highlights representative datasets, modeling strategies, evaluation benchmarks, and the integration of structural and biochemical features. The aim is to provide researchers with a structured and up-to-date reference, guiding the development of more robust, generalizable, and interpretable models for predicting protein stability changes upon mutation. As the field moves forward, the synergy between data-driven AI methods and domain-specific biological knowledge will be key to unlocking deeper understanding and broader applications of protein engineering.

Objective:To investigate the clinical efficacy of moxibustion(Mox)combined with low-dose tadalafil(TAD)in the treatment of diabetes mellitus-induced erectile dysfunction(DMED)with the syndrome of Qi deficiency and blood stasis.Meth-ods:According to the inclusion and exclusion criteria,we selected 90 patients with DMED for this trial and equally randomized them into a Mox,a TAD,and a Mox combined with TAD(Mox+TAD)group to be treated by mild Mox applied to the acupoints Zusanli,Sanyinjiao and Yinlingquan qd alt,oral medication with low-dose TAD at 5 mg per dose qd,and combination of the above two thera-pies,respectively,all for 4 weeks.We obtained from the patients their IIEF-5 scores,traditional Chinese medicine(TCM)symptoms scores,Erectile Hardness Scale(EHS)scores,corpus cavernosal hemodynamic indexes,and the peak systolic velocity(PSV),end diastolic velocity(EDV)and resistance index(RI)of the corpus cavernosal arteries before and after treatment,and compared them among the three groups.Results:The total effectiveness rate was significantly higher in the Mox+TAD(90.0％)than in the Mox(46.7％)and TAD groups(60.0％)(P＜0.05).Compared with the baseline,the IIEF-5 and EHS scores were increased,while the TCM symptoms scores decreased in all the three groups after treatment,more significantly in the Mox+TAD group than in the other two(P＜0.05).And the PSV and RI were remarkably increased,while the EDV decreased(P＜0.05)in all the three groups(P＜0.05)after treatment,with PSV even higher in the Mox+TAD than in the Mox and TAD groups(P＜0.05).Conclusion:Moxi-bustion combined with tadalafil has a definite efficacy and safety for the treatment of DMED,which can effectively improve the erectile function of the patients by increasing penile blood supply,benefiting qi and activating blood circulation.

Objective To explore the value of MRI subtraction technique(ST)for evaluating the efficacy of systemic therapy for advanced hepatocellular carcinoma(HCC)and predicting prognosis after combining with surgery.Methods Totally 35 patients with 39 HCC lesions who received systemic therapy+radical resection were retrospectively collected.Based on preoperative MRI,tumor activity ratio(recorded as tumor activityST)was obtained with ST,while tumor activity value(recorded as tumor activitypathology)was obtained through postoperative pathology,and their correlation was analyzed.The patients were regularly followed up after surgery,and the survival data were recorded.Receiver operating characteristic(ROC)curve was drawn to evaluate the efficacy of tumor activityST for predicting patients'survival status.Then the patients were divided into survival benefit group and no survival benefit group according to the cut-off value,and survival analysis was conducted.Results Tumor activityST was positively correlated with tumor activitypathology(r=0.900,P＜0.001).The median follow-up time was 32.93 months,during which 8 patients died,and the median survival time was 29.9 months.The area under the curve(AUC)of tumor activityST for predicting patients'survival status was 0.67,and the cut-off value was 0.36.Thirty patients with tumor activityST＜0.36 were enrolled in survival benefit group,while 5 patients≥0.36 were collected in no survival benefit group.The overall survival in survival benefit group was longer than that in no survival benefit group(P＜0.001).Conclusion MRI ST could be used to non-invasively evaluate the efficacy of systemic therapy for advanced HCC and predict prognosis after combining with surgery.

Objective:To compare the therapeutic efficacy and safety of local anesthesia and spinal anesthesia for the patients with varicocele(VC)who underwent microsurgical varicocelectomy(MV).Methods:We retrospectively analyzed the data of VC patients who underwent MV treatment at the Andrology Department of the Affiliated Ruikang Hospital of Guangxi University of Chinese Medicine from May 2020 to March 2023.Cases with complete clinical data and follow-up evaluation were selected and divided into a control group(spinal anesthesia)and an observation group(local anesthesia)according to different anesthesia methods.The surgical time(including anesthesia time),visual analogue scale(VAS)score for pain,hospital stay,treatment cost,sperm concentration,for-ward motile sperm rate,and normal sperm morphology rate after three months of surgery,as well as postoperative complications and re-currence rate were compared between the two groups.Results:A total of 107 eligible cases were included,with 56 cases in the con-trol group and 51 cases in the observation group.There was no significant difference in the VAS score for pain during and after four hours of surgery,as well as postoperative complications,and recurrence rate between the two groups(P＞0.05).There was an signif-icant increase in sperm concentration,forward motile sperm rate,and normal sperm morphology rate in both of two groups after three months of surgery(P＜0.05).However,there was no significant difference between the two groups three months after surgery(P＞0.05).The surgical time and hospital stay were shorter than those of the control group(P＜0.05).And the treatment cost in observa-tion group was lower than that of the control group(P＜0.05).Conclusion:Both local anesthesia and lumbar anesthesia for MV treatment of VC have good efficacy and safety.However,patients treated with MV under local anesthesia for VC have obvious advanta-ges in terms of operation time(including anesthesia time),hospital stay,and treatment cost,which is worthy of clinical promotion and application.

Objective To investigate the clinical efficacy and safety of facilitated percutaneous coronary intervention(PCI)with half-dose recombinant staphylokinase(r-SAK)in patients with ST-segment elevation myocardial infarction(STEMI)who are expected to undergo PCI within 120 minutes.Methods From October 2021 to August 2022,a total of 200 STEMI patients in eight centers were included and randomly assigned in a 1﹕1 ratio to either r-SAK group or control group.Patients received loading doses of aspirin and ticagrelor and intravenous heparin and were randomized to receive an intravenous bolus of either 5 mg r-SAK or normal saline prior to PCI.The outcomes were set as ST-segment resolution(STR)at 60-90 minutes after PCI,the proportion and transition of pathological Q waves on the 5th day after PCI,and the proportion of high-sensitivity cardiac troponin T(hs-cTnT)peaking within 12 hours of onset.The safety outcome was major bleeding events defined as Bleeding Academic Research Consortium(BARC)≥type 3 bleeding during hospitalization.Results Compared with the control group,the r-SAK group had a higher proportion of STR≥70%within 60-90 minutes after PCI(58.3%vs.40.3%,P=0.009);a lower proportion of pathological Q waves(59.1%vs.74.1%,P=0.040);a lower rate of Q wave progression(14.8%vs.43.2%,P＜0.001);a higher rate of Q wave disappearance(12.5%vs.3.7%,P=0.027);and a higher proportion of hs-cTnT peaking within 12 hours of symptom onset[31/40(77.5%)vs.17/33(51.5%),P=0.027].Regarding the safety outcome,no significant difference in BARC≥type 3 bleeding was found between the two groups during hospitalization(P＞0.05).Conclusions For STEMI patients who were expected to undergo primary PCI within 120 minutes of symptom onset,the facilitated PCI with half-dose r-SAK significantly increased the proportion of STR≥70%at 60-90 minutes after PCI,reduced the formation of pathological Q waves,and shortened the time to peak hs-cTnT,without increasing the risk of bleeding,which should be an alternative reperfusion strategy worthy of further study.

Aim To evaluate the role of the glucose transporter protein 1(GLUT1)-dependent glycolytic in the attenuation of oxygen-glucose deprivation-reoxygen-ation(OGD/R)injury in HK-2 cells by dexmedetomi-dine(Dex).Methods C57/BL6 mice were random-ly divided into three groups(n=6),namely,sham operation group(Sham group),renal ischemia reper-fusion group(I/R group)and Dex group(I/R+Dex group).Serum creatinine(Cr)and urea nitrogen(BUN)were measured,while the levels of key glyco-lytic enzymes HK2,PFKFB3 and GLUT1 were meas-ured.HK-2 cells were cultured and randomised into seven groups(n=6),which was treated with OGD/R,overexpression or interference with GLUT1,Dex and glycolysis inhibitor 2-DG.CCK-8 and LDH activi-ty were used to detect cellular damage.Glycolysis lev-els were detected by lactate and ECAR.The inflamma-tory level was reflected by qRT-PCR for IL-6 and TNF-α.qRT-PCR and Western blot were performed to de-tect the levels of GLUT1,HK2,and PFKFB3.Results Dex significantly ameliorated kidney injury and HK-2 cell injury(P＜0.05).Dex inhibited the OGD/R-induced rise in lactate and extracellular acidification rate(ECAR),as evidenced by suppression of the ex-pression of GLUT1,HK2 and PFKFB3(P＜0.05).In vitro experiments showed that GLUT1 knockdown sig-nificantly improved OGD/R-induced cellular damage.Lactate,ECAR,glycolysis-related mRNAs and pro-teins were inhibited by GLUT1 knockdown(P＜0.05).Significantly,there were no significant differ-ences in above indexes after Dex treatment based on GLUT1 knockdown.Overexpression of GLUT1 abroga-ted the protective effects of Dex,while reversing the inhibitory effects of Dex on the expression of GLUT1,HK2,and PFKFB3(P＜0.05).Conclusions Dexmedetomidine attenuates OGD/R induced injury in HK-2 cells by inhibiting GLUT1-dependent glycolysis.

Objective To propose a precise suctioning scheme for intravenous drug dispensing based on the vial dispensing robot to enhance the quality of finished infusion.Methods Six kinds of typical representative vial drugs were selected as the research objects,including pantoprazole sodium for injection,papaverine hydrochloride for injection,cefuroxime sodium for injection,Bozhi Glycopeptide Injection,Esomeprazole Sodium for Injection and Methylprednisolone Sodium Succinate for injection.The optimal suction speed was determined by studying the relationship between the size parameters of vials and the suction speed of robot dispensing,which was used to carry out drug dispensing with the vial dispensing robot to verify whether the minimum drug residue could be obtained with the speed.The drug residue was compared with that by manual dispensing.SPSS 24.0 and Excel 2021 were applied to statistical analysis.Results The optimal suction efficiency and minimized drug residue could be got with the depth of the syringe needle hole into the rubber plug(X)less than the height of the rubber plug(H)and the optimal suction speed(Vs)of 7.48 mm/s;the suction efficiency could be ensured without air drawn in when X not less than H and Vs ranging from 10.64 to 39.31 mm/s.The mean values of the drug residue by the robot were all lower than those by manual dispensing,with the differences being statistically significant(P<0.05).Conclusion The proposed scheme can be used for optimizing the parameters of the vial vial dispensing robot to obtain infusion solution with high stability and reliability,which promotes standardization and normalization of intravenous infusion dispensing process.[Chinese Medical Equipment Journal,2025,46(9):45-51]

Due to the inherently complex anatomical structure and physiological charac-teristics of the biliary system, biliary surgery faces considerable challenges in key stages such as preoperative planning, intraoperative identification, and postoperative management. These challenges contribute to a persistently high incidence of complications, severely affecting patient prognosis and overall survival. In recent years, with the continued advancement of interdisciplinary integration between medicine and engineering, intelligent digital technologies, centered on digital medicine, big data, and medical artificial intelligence, have been increasingly applied in the field of biliary surgery, offering new solutions to these longstanding problems. The authors systematically review recent advances and clinical applications of intelligent digital technologies in biliary surgery, and explore the evolution from clinical needs to technological innovation and practical implementation. The aim is to facilitate the transition of biliary surgery from an experience-based approach to an intelligent and precision-driven paradigm, thereby improving surgical safety and patient outcomes.

Pancreatic surgery remains one of the most challenging fields in general surgery due to its complex anatomy and high risk of complications.In recent years,the integration of digital intelligent technologies-such as three-dimensional(3D)reconstruction,fluorescence navigation,augmented/mixed reality(AR/MR),and artificial intelligence(AI)-has provided new strategies for achieving precision,safety,and intelligence in pancreatic surgery.This review summarizes the innovations and clinical applications of these technologies throughout the full perioperative process,including preoperative planning,intraoperative navigation,and postoperative monitoring.Preoperatively,the deep learning-based nnU-Net framework enables high-precision 3D reconstruction of peripancreatic vessels and pancreaticobiliary ducts for individualized surgical planning.Intraoperatively,multimodal image fusion combining indocyanine green fluorescence imaging with AR/MR navigation allows real-time visualization and spatial localization of key anatomical structures,enhancing surgical accuracy and safety.Postoperatively,AI-driven models integrated with wearable sensors and multimodal data support intelligent risk prediction and early intervention for complications.The systematic application of digital intelligence is reshaping the paradigm of pancreatic surgery,driving the transition from experience-based to data-driven and from surgeon-dependent to intelligent decision-making.With continuous algorithmic refinement and accumulation of clinical evidence,digital empowerment will further advance pancreatic surgery toward personalization,precision,and intelligence.

Objective To propose a decision tree-based optimization algorithm to shorten the time spent and to improve the decision-making efficiency for medical device fault diagnosis.Methods Firstly,a binarized fault state space for medical equipment was constructed;secondly,the binarized fault state space was simplified with logical functions;finally,a greedy algorithm was applied to iterative judgment to build a decision tree for medical device fault diagnosis.The algorithm proposed was compared with the greedy algorithm,Gini coefficient method and random forest algorithm in terms of the result accuracy and time cost and the influences of fault probability distribution and time-consuming differences for the algorithms.A case study on gastric lavage machine fault diagnosis was conducted to demonstrate the algorithm.Results The algorithm proposed behaved better than the greedy algorithm,Gini coefficient method and random forest algorithm in lowered time cost and result accuracy.However,the algorithm showed a decrease in accuracy for the logarithmic distribution of the detection method's time consumption,and it was not sensitive to the changes in the failure rate distribution.The gastric lavage machine case study demonstrated that the algorithm was functionally similar to an experienced engineer.Conclusion The proposed algorithm with overall perspective effectively enhances fault judgment efficiency and facilitates rapid diagnostic decision-making in medical equipment maintenance.[Chinese Medical Equipment Journal,2025,46(5):78-83]