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 understand the respiratory protection competency of staff in hospitals.Methods Staff from six hospitals of different levels and characteristics in Beijing were selected,including doctors,nurses,medical technicians,and servicers,to conduct knowledge assessment on respiratory protection competency.According to exposure risks of respiratory infectious diseases,based on actual cases and daily work scenarios,content of respira-tory protection competency assessment was designed from three aspects:identification of respiratory infectious di-seases,transmission routes and corresponding protection requirements,as well as correct selection and use of masks.The assessment included 6,6,and 8 knowledge points respectively,with 20 knowledge points in total,all of which were choice questions.For multiple-choice questions,full marks,partial marks,and no mark were given respective-ly if all options were correct,partial options were correct and without incorrect options,and partial options were correct but with incorrect options.Difficulty and discrimination analyses on question of each knowledge point was conducted based on classical test theory.Results The respiratory protection competency knowledge assessment for 326 staff members at different risk levels in 6 hospitals showed that concerning the 20 knowledge points,more than 60％participants got full marks for 6 points,while the proportion of full marks for other questions was relatively low.Less than 10％participants got full marks for the following 5 knowledge points:types of airborne diseases,types of droplet-borne diseases,conventional measures for the prevention and control of healthcare-associated infec-tion with respiratory infectious diseases,indications for wearing respirators,and indications for wearing medical protective masks.Among the 20 knowledge questions,5,1,and 14 questions were relatively easy,medium,and difficult,respectively;6,1,4,and 9 questions were with discrimination levels of ≥0.4,0.30-0.39,0.20-0.29,and ≤0.19,respectively.Conclusion There is still much room for hospital staff to improve their respiratory protection competency,especially in the recognition of diseases with different transmission routes and the indications for wearing different types of masks.

Fluorescent carbon dots(CDs)were synthesized via a solvothermal method with citric acid and urea as raw materials,and ethylene glycol as reaction solvent.The micromorphology,crystal structure,elemental composition,surface functional group,and optical property of as-synthesized CDs were characterized.The excitation-dependent fluorescence property of CDs was investigated,and the effects of synthesis conditions including reaction temperature,reaction time and raw materials on excitation and emission wavelengths of the CDs were also discussed.Then,a series of CDs-based fluorescent composites were prepared by combining CDs with starch,nano-silica,montmorillonite,kaoline,kieselguhr and magnesium oxide,respectively.Finally,the CDs-starch composites were used for latent fingerprint development on smooth substrates,and the qualitative as well as quantitative evaluation of the contrast,sensitivity and selectivity in fingerprint development were also made.Enhanced development of latent fingerprints was thus achieved by the aid of the excitation-dependent fluorescence property of CDs-starch composite combined with the optical filtering technique,which could decrease the background noise interference to a great extent.Experimental results showed that,the contrast between fingerprint(developing signal)and substrate(background noise)was obvious,exhibiting a strong contrast;the minutiae of papillary ridges were clear,indicating a high sensitivity;the adsorption between CDs-starch composites and fingerprint residues was specific,showing a good selectivity.

A serious of rare earth luminescent micro/nano-materials with various properties were synthesized via chemical method for fluorescent development of latent fingerprints(LFPs).Three evaluation indexes namely contrast,sensitivity and selectivity were introduced to evaluate the effects of LFP development.Quantitative formulas for calculating the contrast,sensitivity and selectivity were further put forward,and a quality evaluation system based on Python was thus established.In addition,the objective evaluation value was finally confirmed to be consistent with the subjective visual judgment.The reproducibility of this evaluation method was finally confirmed.The effects of luminescence intensity and color of developing materials on the contrast,particle size of developing materials on the sensitivity,and micromorphology and surface property of developing materials on the selectivity were discussed in detail.Five effective ways were also proposed to promote the quality of LFP development,such as increasing the luminescence intensity,tuning the luminescence color,decreasing the particle size,adjusting the micromorphology,and modifying the surface property.This quality evaluation system based on Python could evaluate the effects of LFP development objectively,accurately and comprehensively,exhibiting easy operability,high efficiency,sensitive response,accurate and reliable results,and wide applicability,which would provide beneficial references for the reasonable selection of LFP development methods as well as objective evaluation of evidence value.

Near infrared spectroscopy(NIRS)technology combined with chemometrics algorithms has been widely used in quantitative and qualitative analysis of food and medicine.However,traditional chemometrics methods,especially linear classification methods,often yield unsatisfactory results when addressing multi-class classification problems.Convolutional neural network(CNN)is adept at extracting deep-level features from data and suitable for handling non-linear relationships.The modeling performance of CNN depends on the size and diversity of sample,while the collection and preprocessing of NIRS sample data is often time-consuming and labor-intensive.This study proposed a NIRS qualitative analysis method based on data augmentation strategies and CNN.The data augmentation strategy included two steps.Firstly,applying Bootstrap resampling and generative adversarial network(GAN)methods to augment three NIRS datasets(Medicine,coffee and grape).Secondly,combining the original samples(Y)with the Bootstrap augmented samples(B)and GAN augmented samples(G)to obtain three augmented datasets(Y-B,Y-G and Y-B-G).Based on this,a CNN model structure suitable for these datasets was designed,consisting of 2 one-dimensional convolutional layers,1 max-pooling layer,and 1 fully connected layer.The results showed that compared to the optimal models of partial least squares discriminant analysis(PLS-DA),support vector machine(SVM),and back propagation neural network(BP),the CNN model based on Y-B dataset achieved average accuracy improvements of 3.998%,9.364%,and 4.689%for medicine(Binary classification);the CNN model based on the Y-B-G dataset achieved average accuracy improvements of 6.001%,2.004%,and 7.523%for coffee(7-class classification);and the CNN model based on the Y-B dataset achieved average accuracy improvements of 33.408%,51.994%,and 34.378%for grapes(20-class classification).It was evident that the models established based on data augmentation strategies and CNN demonstrated better classification accuracy and generalization performance with different datasets and classification categories.

Objective To detect the expression changes of interleukin-10(IL-10)and transforming growth factor-β1(TGF-β1)during the development of deep vein thrombosis in mice,and to explore the application value of them in thrombus age estimation.Methods The mice in the experimental group were subjected to ligation of inferior vena cava.The mice were sacrificed by excessive anesthesia at 1 d,3 d,5 d,7 d,10 d,14 d and 21 d after ligation,respectively.The inferior vena cava segment with thrombosis was extracted below the ligation point.The mice in the control group were not ligated,and the inferior vena cava segment at the same position as the experimental group was extracted.The ex-pression changes of IL-10 and TGF-β1 were detected by immunohistochemistry(IHC),Western blot-ting and real-time qPCR.Results IHC results revealed that IL-10 was mainly expressed in monocytes in thrombosis and TGF-β1 was mainly expressed in monocytes and fibroblast-like cells in thrombosis.Western blotting and real-time qPCR showed that the relative expression levels of IL-10 and TGF-β1 in each experimental group were higher than those in the control group.The mRNA and protein levels of IL-10 reached the peak at 7 d and 10 d after ligation,respectively.The mRNA expression level at 7 d after ligation was 4.72±0.15 times that of the control group,and the protein expression level at 10 d after ligation was 7.15±0.28 times that of the control group.The mRNA and protein levels of TGF-β1 reached the peak at 10 d and 14 d after ligation,respectively.The mRNA expression level at 10 d after ligation was 2.58±0.14 times that of the control group,and the protein expression level at 14 d after ligation was 4.34±0.19 times that of the control group.Conclusion The expressions of IL-10 and TGF-β1 during the evolution of deep vein thrombosis present time-dependent sequential changes,and the expression levels of IL-10 and TGF-β1 can provide a reference basis for thrombus age estimation.

Objective To investigate the nutritional status of children with Crohn's Disease (CD) at diagnosis and its association with clinical characteristics. Methods A retrospective analysis was performed for the clinical data and nutritional status of 118 children with CD who were admitted to Beijing Children's Hospital,Capital Medical University,from January 2016 to January 2024. A multivariate logistic regression analysis was used to investigate the risk factors for malnutrition. Results A total of 118 children with CD were included,among whom there were 68 boys (57.6％) and 50 girls (42.4％),with a mean age of (11±4) years. Clinical symptoms mainly included recurrent abdominal pain (73.7％,87/118),diarrhea (37.3％,44/118),and hematochezia (32.2％,38/118),and 63.6％ (75/118) of the children had weight loss at diagnosis. The incidence rate of malnutrition was 63.6％ (75/118),and the children with moderate or severe malnutrition accounted for 67％ (50/75). There were 50 children (42.4％) with emaciation,8 (6.8％) with growth retardation,and 9 (7.6％) with overweight or obesity. Measurement of nutritional indices showed a reduction in serum albumin in 83 children (70.3％),anemia in 74 children (62.7％),and a reduction in 25 hydroxyvitamin D in 15 children (60％,15/25). The children with malnutrition had significantly higher disease activity,proportion of children with intestinal stenosis,and erythrocyte sedimentation rate and a significant reduction in serum albumin (P<0.05). The multivariate logistic regression analysis showed that intestinal stenosis was an independent risk factor for malnutrition in children with CD (OR=4.416,P<0.05). Conclusions There is a high incidence rate of malnutrition in children with CD at diagnosis,which is associated with disease activity and disease behavior. The nutritional status of children with CD should be closely monitored.

Objective To evaluate the application effect of broth microdilution(BMD)method in susceptibility testing of multidrug-resistant Mycobacterium tuberculosis(MDR-MTB).Methods The Roche's proportion method and BMD method were adopted in drug susceptibility testing on 108 MDR-MTB strains and 11 non-MDR-MTB strains in Hainan Province.Whole genome sequencing(WGS)was performed on strains with inconsistent results by the above two methods.Results The average time to acquire drug susceptibility testing results by Roche's propor-tional method and BMD method were 28.0 and 8.5 days,respectively.Roche's proportional method showed higher resistance rates to isoniazid(INH),rifampicin(RFP),ethambutol(EMB),kanamycin(KM),and capreomycin(CPM)than BMD method(all P＜0.001).BMD method showed higher resistance rates to protionamide(PTO)and para-aminosalicylic acid(PAS)than Roche's proportional method(both P＜0.001).Taking Roche's proportional method as the gold standard,the sensitivity and specificity of BMD method for testing drug resistance were 50.00％-100％and 95.69％-100％,respectively.Except EMB(87.39％)and INH(94.96％),the consistency rates of the BMD method in testing drug resistance of other drugs were all ≥95.00％.The overall consistency rate between Roche's proportional method and WGS was 76.19％(32/42),while the consistency rate between BMD method and WGS was 23.81％(10/42),difference was statistically significant(x2=23.048,P＜0.001).34 MTB strains showed inconsistent results by two drug susceptibility testing methods.Among the 26 MTB strains that were resis-tant in Roche's proportion method but sensitive in BMD method,22 strains(84.62％)had mutations in relevant re-sistance genes.Among the 11 MTB strains that were sensitive in Roche's proportion method but resistant in BMD method,5 strains(45.45％)had mutations in relevant resistance genes.Conclusion BMD method is an accurate and rapid MDR-MTB susceptibility testing method,but further improvement and optimization are still needed.Drug resistance is closely related to mutations in relevant resistance genes.

Objective To monitor the susceptibility of clinical isolates to antimicrobial agents in tertiary hospitals in major regions of China in 2022.Methods Clinical isolates from 58 hospitals in China were tested for antimicrobial susceptibility using a unified protocol based on disc diffusion method or automated testing systems.Results were interpreted using the 2022 Clinical &Laboratory Standards Institute(CLSI)breakpoints.Results A total of 318 013 clinical isolates were collected from January 1,2022 to December 31,2022,of which 29.5％were gram-positive and 70.5％were gram-negative.The prevalence of methicillin-resistant strains in Staphylococcus aureus,Staphylococcus epidermidis and other coagulase-negative Staphylococcus species(excluding Staphylococcus pseudintermedius and Staphylococcus schleiferi)was 28.3％,76.7％and 77.9％,respectively.Overall,94.0％of MRSA strains were susceptible to trimethoprim-sulfamethoxazole and 90.8％of MRSE strains were susceptible to rifampicin.No vancomycin-resistant strains were found.Enterococcus faecalis showed significantly lower resistance rates to most antimicrobial agents tested than Enterococcus faecium.A few vancomycin-resistant strains were identified in both E.faecalis and E.faecium.The prevalence of penicillin-susceptible Streptococcus pneumoniae was 94.2％in the isolates from children and 95.7％in the isolates from adults.The resistance rate to carbapenems was lower than 13.1％in most Enterobacterales species except for Klebsiella,21.7％-23.1％of which were resistant to carbapenems.Most Enterobacterales isolates were highly susceptible to tigecycline,colistin and polymyxin B,with resistance rates ranging from 0.1％to 13.3％.The prevalence of meropenem-resistant strains decreased from 23.5％in 2019 to 18.0％in 2022 in Pseudomonas aeruginosa,and decreased from 79.0％in 2019 to 72.5％in 2022 in Acinetobacter baumannii.Conclusions The resistance of clinical isolates to the commonly used antimicrobial agents is still increasing in tertiary hospitals.However,the prevalence of important carbapenem-resistant organisms such as carbapenem-resistant K.pneumoniae,P.aeruginosa,and A.baumannii showed a downward trend in recent years.This finding suggests that the strategy of combining antimicrobial resistance surveillance with multidisciplinary concerted action works well in curbing the spread of resistant bacteria.

Objective To investigate the distribution and antimicrobial resistance profiles of the common pathogens isolated from urine from 2015 to 2021 in the CHINET Antimicrobial Resistance Surveillance Program.Methods The bacterial strains were isolated from urine and identified routinely in 51 hospitals across China in the CHINET Antimicrobial Resistance Surveillance Program from 2015 to 2021.Antimicrobial susceptibility was determined by Kirby-Bauer method,automatic microbiological analysis system and E-test according to the unified protocol.Results A total of 261 893 nonduplicate strains were isolated from urine specimen from 2015 to 2021,of which gram-positive bacteria accounted for 23.8％(62 219/261 893),and gram-negative bacteria 76.2％(199 674/261 893).The most common species were E.coli(46.7％),E.faecium(10.4％),K.pneumoniae(9.8％),E.faecalis(8.7％),P.mirabilis(3.5％),P.aeruginosa(3.4％),SS.agalactiae(2.6％),and E.cloacae(2.1％).The strains were more frequently isolated from inpatients versus outpatients and emergency patients,from females versus males,and from adults versus children.The prevalence of ESBLs-producing strains in E.coli,K.pneumoniae and P.mirabilis was 53.2％,52.8％and 37.0％,respectively.The prevalence of carbapenem-resistant strains in E.coli,K.pneumoniae,P.aeruginosa and A.baumannii was 1.7％,18.5％,16.4％,and 40.3％,respectively.Lower than 10％of the E.faecalis isolates were resistant to ampicillin,nitrofurantoin,linezolid,vancomycin,teicoplanin and fosfomycin.More than 90％of the E.faecium isolates were ressitant to ampicillin,levofloxacin and erythromycin.The percentage of strains resistant to vancomycin,linezolid or teicoplanin was＜2％.The E.coli,K.pneumoniae,P.aeruginosa and A.baumannii strains isolated from ICU inpatients showed significantly higher resistance rates than the corresponding strains isolated from outpatients and non-ICU inpatients.Conclusions E.coli,Enterococcus and K.pneumoniae are the most common pathogens in urinary tract infection.The bacterial species and antimicrobial resistance of urinary isolates vary with different populations.More attention should be paid to antimicrobial resistance surveillance and reduce the irrational use of antimicrobial agents.