BackgroundHepatic arterial infusion chemotherapy (HAIC) is increasingly used to treat unresectable hepatocellular carcinoma (HCC), yet some HCC patients remain unresponsive or show disease progression.PurposeThis phase II study (ChiCTR2300077718) aimed to assess the efficacy and safety of balloon-occluded HAIC (bHAIC) based on the FOLFOX (oxaliplatin plus fluorouracil and leucovorin) regimen (bHAIC-FO) in unresectable HCC.Materials and methodsThe present work is an investigator-initiated, multicenter, single-arm, prospective cohort trial. Unresectable HCC patients who had undergone a successful bHAIC procedure were included in the efficacy and safety analysis. After all patients received at least one post-bHAIC follow-up, an interim analysis was conducted in accordance with the predefined protocol. The primary endpoint of this interim analysis was the objective response rate (ORR). Secondary endpoints included changes in alpha-fetoprotein (AFP) levels and the occurrence of adverse events (AEs). Interim data were collected and analyzed using SPSS, version 27.0.Results50 patients received at least one successful bHAIC procedure and were included in the efficacy and safety analysis. The best ORR for 45 evaluable patients was 53.3% according to Response Evaluation Criteria in Solid Tumors 1.1 (RECIST 1.1) criteria, and 82.2% according to modified RECIST (mRECIST) criteria. In the patients population, 36 treatment-naive individuals achieved an 88.9% ORR according to mRECIST criteria, with 15 cases achieving radiological complete response (CR). Overall, 77.3% of patients with elevated AFP levels experienced a reduction of over 50% in their AFP levels. All AEs were manageable by medication or dose modifications.ConclusionbHAIC-FO demonstrates encouraging efficacy and acceptable safety profiles in unresectable HCC treatment, particularly in treatment-naive patients.Clinical trial registrationwww.chictr.org.cn, identifier ChiCTR2300077718

IntroductionColorectal cancer liver metastasis (CRLM) develops in a hepatic niche where tumor cells encounter organ‑specific microbial, metabolic, and immune constraints. The role of intratumoral bacteria in this metastatic ecosystem remains incompletely defined.MethodsWe analyzed tumor and peritumoral tissues from 20 patients with non‑metastatic colorectal cancer (CRC) and 20 patients with CRLM by 5R 16S rDNA sequencing, untargeted metabolomics, and functional assays. Public datasets were used for single‑cell analysis of the tumor immune microenvironment.ResultsPrimary colorectal lesions and paired liver metastases showed distinct microbial structures. Liver metastatic lesions had lower microbial diversity, and 73 genera differed between CRLM primary tumors and liver metastases. Bacillus was enriched in metastatic tumors and was associated with fatty‑acid‑related microbial pathways, serum CEA and CA19‑9 levels, and lipid‑related metabolites that accumulated in liver metastases. Bacillus supernatant increased SW480 and SW620 cell migration and proliferation in vitro, with the strongest activity at a multiplicity of infection (MOI)‑equivalent of 20. In an intrasplenic CRLM model, viable Bacillus and Bacillus supernatant increased metastatic tumor growth, whereas antibiotic intervention reduced this effect. Metabolomic profiling detected enrichment of lipids and lipid‑like molecules, including carnitine‑ and bile‑acid‑related metabolites, in liver metastatic lesions. Single‑cell analysis of public datasets showed remodeling of the tumor immune microenvironment (TIME) and identified APOE⁺ macrophages as lipid‑associated myeloid cells with broad ligand‑receptor communication with epithelial, immune, and stromal compartments.DiscussionTogether, these findings provide evidence that Bacillus promotes metastatic phenotypes and reveal associations among intratumoral Bacillus, lipid‑related metabolic alterations, and APOE+ macrophage‑associated communication signatures in CRLM.

Technical advances and a better understanding of liver tolerance have changed the role of radiotherapy in the treatment of liver cancer. The goal of this retrospective study was to evaluate the outcomes of radiotherapy with curative intent in patients ineligible for surgery and palliative intent in patients with advanced disease. Patients were treated with stereotactic body radiotherapy (SBRT), hypofractionated radiotherapy, or conventionally fractionated radiotherapy. Patients with a solitary liver tumor, no vascular invasion, Child-Pugh class A, and no extrahepatic tumors at the time of radiotherapy were considered for radiotherapy with curative intent, and the remaining patients for radiotherapy with palliative intent. Kaplan-Meier statistics and Cox proportional hazards regression analyses were used to evaluate overall survival. Between 04/2019 and 08/2022, 93 consecutive patients with hepatocellular carcinoma (HCC) were treated external beam radiotherapy in our department. A total of 94.6

Magnetic resonance imaging (MRI) is a commonly used clinical imaging examination characterized by high soft tissue resolution, multiparametric imaging capabilities, and multiplanar reconstruction. It is suitable for the precise diagnosis and dynamic assessment of various organ diseases. MRI plays a crucial role in the diagnosis and surgical decisionmaking for anal fistula, particularly in complex cases. This expert consensus focuses on five key aspects of MRI for anal fistula: indications and contraindications, examination protocols, fistula classification and image interpretation, reporting content, and precautions. The aim is to standardize MRI examination protocols and reporting content for anal fistula, thereby enhancing its clinical utility.

Maturity-onset diabetes of the young (MODY) is a monogenic disease that is often undiagnosed or misdiagnosed in China, leading to inappropriate therapy. In this study, we aimed to determine the pathogenic variants in Chinese patients with clinically suspected MODY. A total of 20 probands from 20 unrelated families with suspected MODY in Taizhou Hospital affiliated to Wenzhou Medical University from 2019 to 2023 were enrolled. Whole-exome sequencing (WES) and WES-based copy number variant (CNV) detection were performed for mutational analysis, and the genotype and phenotype of these affected patients were analyzed. In this cohort, there were 12 males (60.0

In recent years, EUS-guided biliary drainage has been widely implemented in China, and standardization of this procedure has become an urgent matter. As EUS-guided biliary drainage technology and accessories continue to advance, the understanding of its clinical benefits and complications is also evolving. This expert consensus summarizes the current evidence and presents 12 clinical questions and 40 recommendations in the form of questions and answers regarding indications and contradictions, techniques, accessory selection, complications, perioperative management, and learning and training. The goal of this consensus is to assist in decision-making and standardize the treatment process.

Patients worldwide suffer from high-frequency adverse events (AEs). However, the reported rate of severe AEs is significantly below the actual incidence. In particular, patient harm caused by clinical laboratory AEs is generally hidden, indirect, and delayed, and patient safety-oriented quality management models have not been well established. Promoting the recognition of errors, risk control, and safety culture by developing and learning from an AE database could improve patient safety and medical quality. Therefore, this study aimed to develop a patient safety-oriented quality management model by analyzing the risk priority of adverse events (RPAEs) in an AE database and achieve standardization, risk control, and continuous improvement using procedural safety checklists. This study, launched in January 2008, retrospectively examined a multisource AE database of the emergency laboratory at Taizhou Enze Medical Center between 2008 and 2023. This study graded and classified 1,012 AEs from internal and external staff, hospital leaders, patient complaints, and auditors according to severity and content based on the standards of the National Health Commission of the People's Republic of China and the International Organization for Standardization standards for medical laboratories (ISO 22367). The subscore of a specific class of an AE category was obtained by calculating the severity score and frequency score, and the risk score of a specific AE category was calculated by summing subscores of all classes. AE categories in the top 80% of the total risk scores were considered RPAEs, the focus for improvement and quality checks. Among the AEs, 98.62% were reported between 2014 and 2023; 68.67% involved patients and 21.84% caused patient harm. High-risk processes included information inconsistency at sample packet encapsulation and sample receipt, and failure of the pneumatic logistics transmission system (PLTS) in the preanalytical phase; reagent and consumable errors and incorrect results in the analytical phase; delayed reporting of critical values in the postanalytical phase; and delayed turnaround time, incorrect information system settings, and equipment malfunctions in the whole analytical process. Continuous improvements were implemented using quality management tools, such as information systems, Lean management, and process optimization. Key improvements included information consistency checks on sending and receiving samples; monitoring specimen transportation using a PLTS; standardized verification and confirmation of settings or modifications in the information system; early warning regarding equipment malfunction; and visual management of reagents, consumables, and equipment. Using this information, the authors designed a process safety checklist for on-site and immediate assessments and standardization of staff behaviors in key processes to improve patient safety. The authors developed a quality model referred to as the "RPAEs, the root causes, countermeasures, implementation, and safety checklist" model. Moreover, this study presents future directions for quality management in medical laboratories, for China or other countries, such as constructing an indicator system to evaluate the effect of AEs on patient safety.

BackgroundThe differential diagnosis between Tuberculosis (TB) and Non-tuberculous Mycobacteria (NTM) has historically been constrained by the inadequate sensitivity and specificity of current diagnostic methods. Furthermore, distinguishing between Active Tuberculosis (ATB) and Latent Tuberculosis Infection (LTBI) poses significant challenges. This study aims to develop a molecular differentiation system for ATB, LTBI, and NTM by integrating plasma proteomics with multi-dimensional analytical techniques, while also exploring key biomarkers associated with disease progression and treatment response.MethodsUsing label-free quantitative technology, we conducted a plasma proteomics analysis across five groups: ATB, LTBI, NTM, Cured Patients (CPs), and Healthy Donors (HD). Differentially Expressed Proteins (DEPs) were identified through screening (FC > 1.5 or <0.67, P < 0.05), followed by Gene Ontology/KEGG pathway enrichment, STRING interaction network, and Mfuzz dynamic clustering analysis to systematically elucidate molecular characteristics. Experimental data were validated through a multidimensional quality control system (Pearson correlation coefficient, peptide distribution, molecular weight distribution, etc.). Enzyme-linked immunosorbent assay (ELISA) was employed to detect the plasma expression levels of target proteins across the groups and to facilitate comparisons.ResultsThis study identified 1,338 non-redundant proteins across five cohorts. Comparative analysis revealed 142 DEPs across the three comparative groups (ATB, LTBI, and NTM), which were primarily localized in the extracellular domain. Key findings include: 27 DEPs in the ATB-LTBI group, primarily enriched in inflammatory responses (such as A2M, IL-1R2) and epithelial barrier functions (TGM3, KRT3); 69 DEPs in the ATB-NTM group, characterized by significant changes in immunoglobulin light chains (IGLV2-11) and innate immune effector molecules (S100A8); 46 DEPs in the NTM-LTBI group, closely related to lipid metabolism (APOC3) and extracellular matrix remodeling (FN1). KEGG pathway analysis revealed that DEPs in the ATB-LTBI group were enriched in nitrogen metabolism pathways, those in the ATB-NTM group were associated with thyroid hormone synthesis, and the NTM-LTBI group was involved in phagosome function. Dynamic clustering results showed six treatment response modules: Cluster 1/2 (riboflavin metabolism, complement coagulation pathway) were activated post-treatment, Cluster 3/4 (proteasome, cardiac signaling pathway) exhibited partial reversal in expression, and Cluster 5/6 (platelet activation, cytoskeleton) showed delayed regression. Research confirmed 10 differential proteins between the ATB-CPs and ATB-HD groups, including S100A8, LTA4H, and DEFA1B, which constitute a molecular fingerprint specific to ATB. ELISA validation confirmed significantly elevated S100A8 and GPX3 in ATB group, while NTM group showed higher FGB and lower ATRN levels.ConclusionsThis study systematically reveals the plasma proteomic characteristics under infection statuses caused by different mycobacteria. A discrimination framework for ATB/LTBI/NTM was constructed based on disease-specific differential proteins, overcoming the limitations of traditional diagnostic techniques in distinguishing infection states. Through dynamic analysis of six temporal therapeutic modules, the reprogramming patterns of the host protein network during tuberculosis treatment were elucidated. This research lays a multidimensional molecular foundation for the precise typing, personalized treatment, and prognostic evaluation of mycobacterial infections.

BACKGROUND:Patients with COVID-19 often produce multiple autoantibodies that impact immune function. This study aimed to assess changes in immune status and correlation with SARS-CoV-2 infection by analyzing dynamic shifts in patients' antinuclear antibody (ANA) profiles. METHODS:A retrospective analysis was conducted on ANA data and clinical characteristics of 680 patients with novel coronavirus pneumonia admitted to Taizhou Enze Medical Center (Group) between December 7, 2022, and January 31, 2023. The analysis covered three phases: early COVID-19 (within one year before admission, T1), COVID-19 phase (during hospitalization, T2), and late COVID-19 (within one year after discharge, T3). ANA quantification was primarily performed using indirect immunofluorescence, and the magnetic stripe immunofluorescence luminescence method was employed to detect the ANA profile (ENA), including anti-dsDNA, nucleosome, Sm, SS-A/Ro52kD, SS-A/Ro60kD, SS-B/La, PCNA, AMA M2, Scl-70, and Jo-1. RESULTS:During the T2 phase, 680 patients were analyzed. The positive rate of the ANA test was 35%. The proportion of autoimmune diseases (AID) in ANA-positive patients was higher than in ANA-negative patients (22%vs.7%). The ANA-positive group with AID showed higher ANA titers compared to the ANA-positive group without AID. During the follow-up one year before and after SARS-CoV-2 infection, in the T1-T2 group, there were two cases of ANA changing from negative to positive (one with AID, one without AID). The positive intensity of ANA increased by 15.6% and decreased by 20%. In the T2-T3 group, the positive intensity of ANA increased by 3.3% and decreased by 33.3%. Followed up of 7 patients with high ANA titers in T2 phase, among whom 5 cases did not support AID from the perspective of diagnosis and medication, and 2 cases were diagnosed with SLE after being infected with SARS-CoV-2. CONCLUSIONS:SARS-CoV-2 infection induces overactivation of the immune system, significantly impacting patients with autoimmune diseases. For patients without autoimmune diseases, ANA produced due to COVID-19 does not persist. Some COVID-19 patients may trigger their own immune system response.

Visual based immunoassay employs both human eye interpretation and machine-vision (smartphone imaging) quantification for equipment-independent diagnosis of biomarkers, fitting a variety of scenarios such as home-testing. Here we use hue-recognition strategy to create vivid and sensitive color changes against C-reactive protein (CRP) for accurate visual analysis. The red-to-green tonality gradient presentation was realized by ratiometric fluorescent probe using dual-emissive quantum dots (QDs) structure. The sandwich immuno-capturing process of antigen was integrated on ELISA plate, with the signal amplified and transduced by silver particle labels etching and interaction with the ratiometric QDs probe. The narrow-emissive QDs with selective response to released Ag+ ions produced high color fidelity and prominent hue variation towards analyte concentration. This portable immunoassay enabled eye-perception of 5 specific CRP concentration points (0, 5, 50, 200, 500 ng/mL) and 4 concentration intervals (0-5, 5-50, 50-200, 200-500 ng/mL) with reading accuracy of 97.5 % and 96.4 %, respectively, which is superior against traditional lightness-gradient based reading mode. With the aid of smartphone imaging and software color-analyzing, an elaborated quantification of CRP was achieved via color information. The broad CRP responsive range (0-500 ng/mL), low limit of detection (0.062 ng/mL) and high specificity against protein substrates allowed a robust point-of-care CRP clinical diagnosis application.

BACKGROUND:Patients with benign gallbladder diseases (BGBD) in China are characterized by prolonged illness duration, recurrent acute inflammation and poor medical compliance, all increase difficulty of laparoscopic cholecystectomy (LC). Complications including biliary injury, residual stones and intraoperative iatrogenic rupture (IIR) of the gallbladder wall (GBW) resulting in incidental gallbladder cancer (IGBC) exposure all warrant vigilance. This study aims to evaluate whether intraoperative ultrasound (IOUS) could enhance safety of crucial steps of LC. MATERIALS AND METHODS:Patients enrolled in this single-center randomized controlled trial were initially diagnosed with BGBD and randomly assigned to two groups at a 1:1 ratio, based on whether IOUS scanning was conducted during LC. The primary aim was to assess whether IOUS could: (1) Identify adherent gastrointestinal tracts within the surgical field; (2) diagnose cystic duct (CD) stones; and (3) guide the selection of appropriate dissection approaches of GBW to prevent IIR according to the modified Gallbladder Reporting and Data System (GB-RADS). RESULTS:A total of 152 patients were enrolled in this trial. IOUS helped in differentiating CD stones and adhered gastrointestinal tissues within the surgical area. IOUS effectively reduced IIR (3.95% vs. 17.11%, p = 0.017) because it clearly demonstrated morphology of GBW layers under different levels of inflammation. Additionally, among patients classified as GB-RADS 2B and 2C, the proportion of cases with atypical hyperplasia was notably higher than those with GB-RADS 2A and 1 (p = 0.002). CONCLUSIONS:IOUS plays a pivotal role in enhancing the safety of critical steps during LC surgeries for selected complex cases. Integrating IOUS with GB-RADS is of great value in guiding the selection of safe GBW approaches and preventing IIR in cases with canceration risk.

Background:Post-hepatectomy liver failure (PHLF) is the leading cause of morbidity and mortality following major hepatectomy. Existing prediction models inadequately capture the dynamic liver regeneration and perioperative changes, limiting their predictive accuracy. We aimed to develop a machine learning (ML) modelling system (PILOT architecture) integrating liver regeneration-associated biomarkers with time-phased perioperative data for PHLF prediction. Methods:This retrospective multicentre study included 1071 patients undergoing major hepatectomy at three centres (2019-2024), divided into training (n = 623) and two external validation cohorts (n = 206 and 242). Fifty-five perioperative variables, including novel liver regeneration-associated biomarkers (GATA3, RAMP2, VEGFA, PEDF), were categorised into three time-phased datasets (preoperative, intraoperative, postoperative). Thirteen ML algorithms were evaluated across these datasets, with gradient-based feature reduction strategies applied to optimise the PILOT models. This study is registered with ClinicalTrials.gov (NCT05779098). Findings:PILOT-Pre, PILOT-Intra (LightGBM with 10 and 15 features, respectively), and PILOT-Post (XGBoost with 20 features) models showed superior discrimination in training (AUCs: 0.754 [95% CI: 0.717-0.790], 0.787 [0.728-0.846], 0.904 [0.883-0.924]) and validation cohorts (AUCs: 0.740-0.895) compared to traditional models (AUCs: 0.502-0.644; all P < 0.050). A risk-stratification framework integrating PILOT-Pre and PILOT-Intra predictions achieved a class-specific precision of 94.4%-96.6% for PHLF events in the consensus high-risk group and 92.1%-95.5% for non-PHLF events in the consensus low-risk populations. SHAP analysis revealed that serum phosphorus levels >2.4 mg/dL on postoperative day 3, liver RAMP2-GATA3 ratios <10.1, and serum PEDF-VEGFA indices >4.9 were associated with an increased predicted PHLF risk. Interpretation:The PILOT architecture integrates liver regeneration-associated biomarkers with time-phased data to accurately predict PHLF within the first 6 h postoperatively. Based on consistency analysis of predictions of PILOT-Pre and PILOT-Intra models, this framework enables early risk stratification, thereby providing a practical tool for personalised perioperative management. Funding:This research was funded by the projects from National Natural Science Foundation of China (82403243), Program for National Postdoctoral Researchers Funding of China (GZC20231943), and Shanghai Municipal Commission of Science and Technology (23Y11905900).

Additive manufacturing (AM) has gained significant traction in aerospace, automobile and medical industries for its capabilities of producing lightweight and customized structures. It is crucial to understand process-structure-property relationships and predict fatigue life of AM built components for engineering reliability. Machine learning (ML) which has become increasingly popular in engineering design and optimization, features in powerful prediction abilities but its performance heavily relies on the data quality and size. Data scarcity of AM built components leads to overfitting and poor accuracy of traditional deterministic ML models and they fail to capture AM material uncertainties. This paper introduces a method combining hierarchical Bayesian models with Bayesian Neural Networks (BNNs) to address these challenges. The approach generates synthetic data that mirrors statistical properties of the original dataset, supplementing ML prediction reliability. BNNs excel deterministic models with limited data while maintaining accuracy and allow for prediction updates as new experimental data becomes available through Bayesian inference. By integrating defect characteristics, the method also provides a probabilistic framework for fatigue life uncertainty quantification. Predictive results demonstrate the potential of this approach in material behavior analysis, offering data-driven fatigue life assessment with uncertainty quantification, and ultimately improving the reliability of AM built components in critical industrial applications.

OBJECTIVES:Bloodstream infections are associated with high morbidity and mortality in patients in the intensive care unit (ICU), necessitating rapid pathogen identification. This study evaluates the diagnostic performance and clinical utility of droplet digital PCR (ddPCR) in patients in the ICU with suspected bloodstream infection. METHODS:This retrospective, single-centre study included 101 adults with suspected bloodstream infection in the ICU. 303 ddPCR samples were collected at three time points after ICU admission: 0-1 (T1), 3-4 (T2), and 6-8 days (T3). ddPCR was compared with blood culture collected at T1 for detection performance, and compared the distributions, copy numbers, and clinical concordance of pathogens across three time points. RESULTS:ddPCR had a 78.2% detection rate versus 32.7% for blood culture and showed 86.0% sensitivity, and 63.6% specificity. Accuracy decreased with delayed sampling (T1: 57.4%, T2: 52.5%, T3: 40.6%). Clinical concordance was >80% for Klebsiella pneumoniae, Escherichia coli, and Staphylococcus aureus; >50% for Pseudomonas aeruginosa and Acinetobacter baumannii; but <40% for Enterococcus and Streptococcus spp. A cut-off of 1 copy/µL for S. aureus, 2 copies/µL for K. pneumoniae, P. aeruginosa, and 5 copies/µL for E. coli, A. baumannii, improved bloodstream infection diagnosis. The detection of multiple pathogens was highest across three time points (T1: 28.7%, T2: 17.8%, T3: 19.8%). 10 in 22 patients with persistently negative tests confirmed to be true negatives. The copy numbers of 28 cases with persistently same pathogen tested correlated dynamically with procalcitonin and C-reactive protein levels. CONCLUSION:ddPCR improves an etiological diagnosis in bloodstream infection, enabling rapid quantification, dynamic monitoring, and better antibiotic stewardship.

Objectives : Elizabethkingia spp. infections pose a major threat to human health with high mortality. This study aimed to further understand its detection status, co-detection patterns, pathogenicity, and antimicrobial resistance in lower respiratory tract infections (LRTI) through metagenomic high throughput sequencing (mNGS)-based real-world research. Design and methods : We retrospectively analyzed 105 LRTI patients positive for Elizabethkingia spp. by mNGS from July 2021 to February 2025. Pathogen profiles, antimicrobial management, and outcomes were reviewed via electronic medical records. Results : mNGS detection rates for Elizabethkingia spp. in respiratory samples were 21.5% in General Intensive Care Unit (GICU) and 11.1% in Emergency Intensive Care Unit (EICU), more sensitive than culture. Polymicrobial co-detection was ubiquitous (99%), indicating a diverse polymicrobial community. Clinical isolates exhibited variable susceptibility (74%-100%) to trimethoprim-sulfamethoxazole, ciprofloxacin, levofloxacin, doxycycline, minocycline, rifampicin, and azithromycin. Patients receiving targeted antimicrobial therapy based on mNGS indicators (Stringent Map Read Number (SMRN) rank ≤2, normalized SMRN (nSMRN) ≥1000, or SMRN percentage ≥25%) had significantly higher effective treatment rates. Conclusions : Elizabethkingia spp. detection rates in ICU respiratory samples are high, frequently complicated by polymicrobial co-detection. Lack of targeted therapy is a key factor in treatment failure. mNGS-derived indicators and local susceptibility databases are essential for guiding effective intervention.

Hepatocellular carcinoma (HCC) is a highly aggressive and heterogeneous malignancy, in which natural killer (NK) cells play a crucial role in tumor progression and immune surveillance. This study aimed to characterize the transcriptomic landscape of NK cell-associated genes (NAGs) and explore their associations with clinical outcomes and therapeutic responses in HCC. Using transcriptomic data from The Cancer Genome Atlas (TCGA), we identified key NAGs through comprehensive statistical analyses. Patients were stratified into distinct risk groups based on NAG expression profiles. Low-risk patients demonstrated better survival, higher immune infiltration, and greater predicted sensitivity to immunotherapy, whereas high-risk patients were associated with reduced chemotherapy responsiveness. These findings contribute to a deeper understanding of the immunogenomic features of HCC and provide a basis for developing personalized therapeutic approaches centered on NK cell-related mechanisms.

This study investigated host responses to long COVID by following up with 89 of the original 144 cohorts for 1-year (N = 73) and 2-year visits (N = 57). Pulmonary long COVID, characterized by fibrous stripes, was observed in 8.7% and 17.8% of patients at the 1-year and 2-year revisits, respectively, while renal long COVID was present in 15.2% and 23.9% of patients, respectively. Pulmonary and renal long COVID at 1-year revisit was predicted using a machine learning model based on clinical and multi-omics data collected during the first month of the disease with an accuracy of 87.5%. Proteomics revealed that lung fibrous stripes were associated with consistent down-regulation of surfactant-associated protein B in the sera, while renal long COVID could be linked to the inhibition of urinary protein expression. This study provides a longitudinal view of the clinical and molecular landscape of COVID-19 and presents a predictive model for pulmonary and renal long COVID.