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.

AIM: To quantify early changes of macular capillary parameters in type 2 diabetic patients using optical coherence tomography angiography(OCTA).METHODS: Retrospective case study. A total of 49 healthy subjects, 52 diabetic patients without retinopathy(noDR)patients, and 43 mild nonproliferative diabetic retinopathy(mNPDR)patients were recruited. Capillary perfusion density, vessel length density(VLD), and average vessel diameter(AVD)were calculated from macular OCTA images(3 mm×3 mm)of the superficial capillary plexus after segmenting large vessels and the deep capillary plexus. Parameters were compared among control subjects, noDR, and mNPDR patients. The area under the receiver operating characteristic curve estimated the abilities of these parameters to detect early changes of retinal microvascular networks.RESULTS: Significant differences were found in the VLD and AVD among the three groups(P<0.001). Compared with the control group, the noDR group had significantly higher AVD(P<0.05). VLD of both layers in patients of mNPDR group was significant decreased compared with that of noDR group(all P<0.01). Deep AVD had a higher area under the curve(AUC)of 0.796 than other parameters to discriminate the noDR group from the healthy group. Deep AVD had the highest AUC of 0.920, followed by that of the deep VLD(AUC=0.899)to discriminate the mNPDR group from the healthy group.CONCLUSIONS: NoDR patients had wider AVD than healthy individuals and longer VLD than mNPDR patients in both layers. When compared with healthy individuals, deep AVD had a stronger ability than other parameters to detect early retinal capillary impairments in noDR patients.

Objective To explore the causal relationship between gastro esophageal reflux disease(GERD)and chronic obstructive pulmonary disease(COPD)by a two-sample Mendelian randomization method.Methods From the summary data of genome-wide association studies(GWAS),single nucleotide polymorphisms(SNP)associated with GERD were selected as genetic instrumental variables.The primary analysis employed the inverse variance weighted(IVW)method for estimating causal effects,with MR-Egger,weighted median,simple mode,and weighted mode methods as supplementary approaches.Sensitivity analyses were conducted to assess the robustness of the results,including Cochran's Q test,MR-Egger intercept,and leave-one-out analysis.Results There might be an association between GERD and an increased risk of COPD[IVW:odds ratio(OR)=1.80,95％confidence interval(CI)1.55-2.08,P＜0.01].Furthermore,no evidence of horizontal pleiotropy and heterogeneity was observed through MR-Egger intercept,Cochran's Q test,and leave-one-out analysis.Conclusion GERD can increase the risk of COPD,and aggressive control of GERD is recommended to reduce the incidence of COPD.

Objective:To report the results of national surveillance on the distribution and antimicrobial resistance profile of clinical Gram-negative bacteria isolates from bloodstream infections in China in 2022.Methods:The clinical isolates of Gram-negative bacteria from blood cultures in member hospitals of national bloodstream infection Bacterial Resistant Investigation Collaborative System（BRICS）were collected during January 2022 to December 2022. Antibiotic susceptibility tests were conducted by agar dilution or broth dilution methods recommended by Clinical and Laboratory Standards Institute（CLSI）. WHONET 5.6 and SPSS 25.0 software were used to analyze the data.Results:During the study period，9 035 strains of Gram-negative bacteria were collected from 51 hospitals，of which 7 895（87.4%）were Enterobacteriaceae and 1 140（12.6%）were non-fermenting bacteria. The top 5 bacterial species were Escherichia coli（ n=4 510，49.9%）， Klebsiella pneumoniae（ n=2 340，25.9%）， Pseudomonas aeruginosa（ n=534，5.9%）， Acinetobacter baumannii complex（ n=405，4.5%）and Enterobacter cloacae（ n=327，3.6%）. The ESBLs-producing rates in Escherichia coli， Klebsiella pneumoniae and Proteus spp. were 47.1%（2 095/4 452），21.0%（427/2 033）and 41.1%（58/141），respectively. The prevalence of carbapenem-resistant Escherichia coli（CREC）and carbapenem-resistant Klebsiella pneumoniae（CRKP）were 1.3%（58/4 510）and 13.1%（307/2 340）；62.1%（36/58）and 9.8%（30/307）of CREC and CRKP were resistant to ceftazidime/avibactam combination，respectively. The prevalence of carbapenem-resistant Acinetobacter baumannii（CRAB）complex was 59.5%（241/405），while less than 5% of Acinetobacter baumannii complex was resistant to tigecycline and polymyxin B. The prevalence of carbapenem-resistant Pseudomonas aeruginosa（CRPA）was 18.4%（98/534）. There were differences in the composition ratio of Gram-negative bacteria in bloodstream infections and the prevalence of main Gram-negative bacteria resistance among different regions，with statistically significant differences in the prevalence of CRKP and CRPA（ χ2=20.489 and 20.252， P<0.001）. The prevalence of CREC，CRKP，CRPA，CRAB，ESBLs-producing Escherichia coli and Klebsiella pneumoniae were higher in provinicial hospitals than those in municipal hospitals（ χ2=11.953，81.183，10.404，5.915，12.415 and 6.459， P<0.01 or <0.05），while the prevalence of CRPA was higher in economically developed regions（per capita GDP ≥ 92 059 Yuan）than that in economically less-developed regions（per capita GDP <92 059 Yuan）（ χ2=6.240， P=0.012）. Conclusions:The proportion of Gram-negative bacteria in bloodstream infections shows an increasing trend，and Escherichia coli is ranked in the top，while the trend of CRKP decreases continuously with time. Decreasing trends are noted in ESBLs-producing Escherichia coli and Klebsiella pneumoniae. Low prevalence of carbapenem resistance in Escherichia coli and high prevalence in CRAB complex have been observed. The composition ratio and antibacterial spectrum of bloodstream infections in different regions of China are slightly different，and the proportion of main drug resistant bacteria in provincial hospitals is higher than those in municipal hospitals.

Objective:To analyze the value of the modified 5-factor frailty index in assessing postoperative complications and mortality in elderly hip fracture patients.Methods:In this retrospective study, clinical data were collected of hip fracture patients aged 60 years and above surgically treated at Beijing Luhe Hospital affiliated to Capital Medical University between January 2015 and December 2019.Patients' group assignment was based on whether the modified frailty index score was ≤1 or ≥2, and a post-surgery follow-up was conducted for survival at 30 days, 1 year, 2 years, and 4 years, which was analyzed by the Kaplan-Meier method.Multivariate Cox regression analysis was used to identify factors affecting death in elderly patients.Results:A total of 1 208 patients were included, with 890 in the group with the index score ≤1 and 318 in the group with the index score ≥2.There was no difference in mortality at 30 days(1.6% or 14/890 vs.1.9% or 6/318, P=0.707), 1-year(11.3% or 99/874 vs.11.6% or 36/310, P=0.917), 2-years(19.7% or 168/852 vs.24.3% or 73/300, P=0.099)and 4-years(44.0% or 238/541 vs.51.5% or 106/206, P=0.071). The incidence of postoperative complications in the group with the score ≥2 was higher(14.8% or 47/318 vs.9.7% or 86/890, P=0.012), including the incidence of stroke(6.3% or 20/318 vs.1.8% or 16/890, P<0.001)and the incidence of postoperative pneumonia(6.0% or 19/318 vs.3.1% or 28/890, P=0.029), and the differences were statistically significant.Multivariate Cox regression analysis showed that age, being female, the Charlson comorbidity index score and low hemoglobin at admission were risk factors for 1-year, 2-year and 4-year mortality post-surgery(all P<0.05), while the modified frailty index score had no correlation with postoperative mortality. Conclusions:A modified frailty index ≥2 is predictive of increased risk of postoperative pneumonia and stroke in patients with hip fractures, but is not correlated with the risk of postoperative mortality.

Objectives:To investigate the impact of baseline non-high-density lipoprotein cholesterol(non-HDL-C)levels on new-onset cardiovascular disease(CVD)in postmenopausal women. Methods:This prospective cohort study selected 8 893 postmenopausal women who participated from 2006 to 2018 employee health examination of Kailuan Group and had complete total cholesterol(TC)and HDL-C data and no history of CVD.Participants were followed up to 31 December,2021.The primary endpoint was the occurrence of CVD or death.According to the Chinese Lipid Management Guidelines(2023),the participants were divided into non-HDL-C<4.1 mmol/L group(n=6 079),4.1 mmol/L≤non-HDL-C<4.9 mmol/L group(n=1 824)and non-HDL-C≥4.9 mmol/L group(n=990).The cumulative incidence of CVD in different groups of non-HDL-C levels was calculated using the Kaplan-Meier method and tested by log-rank analysis.Multivariate Cox regression model was used to analyze the effects of different non-HDL-C levels on CVD. Results:The mean follow-up time was(10.78±4.48)years,the cumulative incidence of CVD in different non-HDL-C level groups was 1.82%,3.24%and 2.89%,respectively.Kaplan-Meier survival curve showed a statistically significant difference in cumulative incidence among the three groups(log-rank P<0.0001).The results of Cox regression analysis showed that after adjusting for confounding factors such as age and sex,the HR(95%CI)values for CVD in the 4.1≤non-HDL-C<4.9 mmol/L group and the non-HDL-C≥4.9 mmol/L group were 1.40(1.13-1.74)and 1.35(1.03-1.78),respectively. Conclusions:High non-HDL-C levels are an independent risk factor for new-onset CVD in postmenopausal women.

Ankylosing spondylitis (AS) combined with lower cervical fracture is often categorized into unstable fracture, with a high incidence of neurological injury and a high rate of disability and morbidity. As factors such as shoulder occlusion may affect the accuracy of X-ray imaging diagnosis, it is often easily misdiagnosed at the primary diagnosis. Non-operative treatment has complications such as bone nonunion and the possibility of secondary neurological damage, while the timing, access and choice of surgical treatment are still controversial. Currently, there are no clinical practice guidelines for the treatment of AS combined with lower cervical fracture with or without dislocation. To this end, the Spinal Trauma Group of Orthopedics Branch of Chinese Medical Doctor Association organized experts to formulate Clinical guidelines for the treatment of ankylosing spondylitis combined with lower cervical fracture in adults ( version 2024) in accordance with the principles of evidence-based medicine, scientificity and practicality, in which 11 recommendations were put forward in terms of the diagnosis, imaging evaluation, typing and treatment, etc, to provide guidance for the diagnosis and treatment of AS combined with lower cervical fracture.

Objective:To determine incidence of seasonal influenza virus infection in the community and to analyze the factors influencing seasonal influenza virus infection.Methods:This study recruited residents aged 6-59 years to build a cohort in 15 villages/streets in Macheng city in November 2019. Meanwhile, a cross-sectional baseline survey was conducted immediately to collect sera, information on demographics and child protection knowledge, behaviors, as well as attitudes using a questionnaire from the participants enrolled in the cohort (i.e., before the influenza epidemic season). In July 2020, a cross-sectional follow-up survey was conducted to collect sera once again (i.e., after the influenza season). Paired sera from the two cross-sectional surveys were tested for influenza virus-specific antibodies by hemagglutination inhibition (HI) test or micro-neutralization (MN) test using a circulating representative strain of each subtype/lineage of influenza virus as the test antigen. The infections with influenza virus subtype/lineage was confirmed if there was a four-fold or more increase in titers of antibodies against circulating representative strain of the subtype/lineage of influenza virus. Factors influencing infection with influenza A (H3N2) and B/Victoria viruses were analyzed using univariable and multivariable logistic regression.Results:In November 2019, 800 study participants were enrolled in the cohort, including 340 children aged 6-17 years and 460 adults aged 18-59 years; 605 study participants (including 224 children and 381 adults) were followed up in July 2020 and their paired sera were obtained before and after the influenza season. 25.3% (153/605) of the participants were confirmed to be infected with at least one subtype/lineage of seasonal influenza virus by HI and MN tests. The overall incidence of influenza viruses of all subtypes/lineages in children was 44.2% (95% CI: 37.6%-50.8%) which was significantly higher than the incidence of 14.1% in adults (95% CI: 10.7%-17.7%). Children had the highest incidence of influenza A (H3N2) virus infection, followed by B/Victoria. MN or HI antibody titers in A (H3N2)[ OR=0.88 (95% CI: 0.84-0.93)] and B/Victoria[ OR=0.97 (95% CI: 0.95-0.99)] before the influenza season were significantly associated with whether children were infected with that subtype/lineage of influenza virus. Conclusions:The residents aged 6-59 years in Macheng city had a substantial incidence of seasonal influenza virus infection during the influenza season from winter 2019 to spring 2020. Notably, almost half of children aged 6-17 years have been infected with seasonal influenza virus. Higher titers of HI/MN antibodies against seasonal influenza virus before the influenza season would be likely to reduce the risk of infection with influenza A (H3N2) and B/Victoria.

Objective:To report the results of national surveillance on the distribution and antimicrobial resistance profile of clinical Gram-positive bacteria isolates from bloodstream infections in China in 2022.Methods:The clinical isolates of Gram-positive bacteria from blood cultures in member hospitals of National Bloodstream Infection Bacterial Resistant Investigation Collaborative System（BRICS）were collected during January 2022 to December 2022. Antibiotic susceptibility tests were conducted by agar dilution or broth dilution methods recommended by Clinical and Laboratory Standards Institute（CLSI）. WHONET 5.6 and SPSS 25.0 software were used to analyze the data.Results:A total of 3 163 strains of Gram-positive pathogens were collected from 51 member units，and the top five bacteria were Staphylococcus aureus（ n=1 147，36.3%），coagulase-negative Staphylococci（ n=928，29.3%）， Enterococcus faecalis（ n=369，11.7%）， Enterococcus faecium（ n=296，9.4%）and alpha-hemolyticus Streptococci（ n=192，6.1%）. The detection rates of methicillin-resistant Staphylococcus aureus（MRSA）and methicillin-resistant coagulase-negative Staphylococci（MRCNS）were 26.4%（303/1 147）and 66.7%（619/928），respectively. No glycopeptide and daptomycin-resistant Staphylococci were detected. The sensitivity rates of Staphylococcus aureus to cefpirome，rifampin，compound sulfamethoxazole，linezolid，minocycline and tigecycline were all >95.0%. Enterococcus faecium was more prevalent than Enterococcus faecalis. The resistance rates of Enterococcus faecium to vancomycin and teicoplanin were both 0.5%（2/369），and no vancomycin-resistant Enterococcus faecium was detected. The detection rate of MRSA in southern China was significantly lower than that in other regions（ χ2=14.578， P=0.002），while the detection rate of MRCNS in northern China was significantly higher than that in other regions（ χ2=15.195， P=0.002）. The detection rates of MRSA and MRCNS in provincial hospitals were higher than those in municipal hospitals（ χ2=13.519 and 12.136， P<0.001）. The detection rates of MRSA and MRCNS in economically more advanced regions（per capita GDP≥92 059 Yuan in 2022）were higher than those in economically less advanced regions（per capita GDP<92 059 Yuan）（ χ2=9.969 and 7.606， P=0.002和0.006）. Conclusions:Among the Gram-positive pathogens causing bloodstream infections in China， Staphylococci is the most common while the MRSA incidence decreases continuously with time；the detection rate of Enterococcus faecium exceeds that of Enterococcus faecalis. The overall prevalence of vancomycin-resistant Enterococci is still at a low level. The composition ratio of Gram-positive pathogens and resistant profiles varies slightly across regions of China，with the prevalence of MRSA and MRCNS being more pronounced in provincial hospitals and areas with a per capita GDP≥92 059 yuan.

Objective:To identify the relationship between the CT arterial radiomics score and the treatment response to neoadjuvant therapy for pancreatic cancer.Methods:The clinical data of 243 pancreatic cancer patients who received surgical resection after neo-adjuvant therapy in the First Affiliated Hospital of Naval Medical University from March 2017 to March 2023 were retrospectively analyzed. Based on the tumor regression grade (TRG), the patients were divided into good response group (TRG 0-1, n=30) and non-good response group (TRG 2-3, n=213). The clinical, radiological and pathological features were compared between two groups. Fully-automated segmentation tool was used for segmenting the arterial CT scan of pancreatic tumor before and after treatment. Python package was applied to extract the radiomics features of tumors after segmentation and the extracted features were reduced and chosen using the least absolute shrinkage and selection operator (Lasso) logistic regression algorithm. Lasso logistic regression formula was applied to calculate the arterial radiomics score. Univariate and multivariate logistic regression models were used to analyze the association between arterial radiomics score and treatment response to neoadjucant therapy. Receiver operating-characteristics (ROC) curve was drawn and area under curve (AUC), specificity, sensitivity and accuracy for evaluating the treatment response were calculated. The clinical usefulness of arterial radiomics score for diagnosing the response of neoadjuvant treatment for pancreatic cancer were determined by decision curve analysis (DCA) . Results:A total of 330 arterial radiomics CT features were obtained, and 9-selected arterial phase features associated with treatment response were determined after being reduced by the Lasso logistic regression algorithm. Univariate analysis showed that the arterial radiomics score, three-dimensional diameter after neoadjuvant therapy, pancreatic contour, T stage, N stage, Peri-pancreatic nerve invasion, lymph-vascular space invasion (LVSI) and invasion of duodenum were all associated with treatment response (all P value <0.05). Multivariate logistic regression analyses confirmed that arterial radiomics score was obviously associated with the neoadjuvant treatment response ( P<0.001). At the cut-off value of 1.93, AUC of the arterial radiomics score for diagnosing neoadjuvant treatment response was 0.92, and the specificity, sensitivity and accuracy was 86.7%, 84.5% and 84.8%. DCA demonstrated that when the percentage for predicting the treatment response by using the arterial radiomics score was >0.2, the patients could benefit from the application of arterial radiomics score for evaluating neoadjuvant therapy response. Conclusions:The arterial radiomics score was strongly correlated with the neoadjuvant treatment response of pancreatic cancer, and can accurately predict neoadjuant treatment efficacy.