Browse Articles

Retraction: The Effects of Matrix Metalloproteinase-9 on Dairy Goat Mastitis and Cell Survival of Goat Mammary Epithelial Cells

Abstract TH933: Age-Related Differences in Uptake of Guideline-Directed Therapy for HFpEF During Hospitalization

Background: Guideline-directed medical therapy (GDMT) is a cornerstone of heart failure (HF) management. However, older adults often receive suboptimal treatment due to comorbidities, polypharmacy, or therapeutic caution. Hospitalization represents a key opportunity to optimize GDMT. This study examined whether older adults hospitalized with HFpEF receive GDMT at rates comparable to younger adults. Methods: We conducted a retrospective cohort study of adults hospitalized with HF and ejection fraction >40% at the University of Rochester Medical Center over a 24-month period. Prescription data at admission and discharge were obtained from the Get With The Guidelines–Heart Failure registry. Patients with incomplete data or contraindications to any GDMT class were excluded. Participants were stratified by age: younger (<65 years) and older (≥65 years). Within-group changes were assessed using McNemar’s tests and between-group comparisons were performed using χ 2 tests. Results: Of 1,190 patients screened, 724 met inclusion criteria; 69.2% were aged ≥65 years. From admission to discharge, both age groups demonstrated significant increases in prescriptions for SGLT2 inhibitors and aldosterone antagonists (both p < 0.001). In older adults, prescriptions for ACE inhibitors and beta-blockers also increased significantly (both p < 0.01), whereas these changes were not observed in younger adults ( p = 0.239 and p = 0.066, respectively). ARB and ARNI use remained unchanged across both groups. At admission, aldosterone antagonist and ARNI use differed by age, but only the difference in aldosterone antagonist prescriptions persisted at discharge. Conclusions: Hospitalization was associated with substantial increases in GDMT utilization among patients with HFpEF, particularly for SGLT2 inhibitors and aldosterone antagonists. Nonetheless, age-related differences in prescribing persisted, underscoring ongoing disparities in care for older adults with HFpEF.

Abstract TH922: Dietary Conjugated Linoleic Acid Shapes the Host Plasma Metabolome and Gut Microbiome and Is Associated with Coronary Heart Disease in U.S. Individuals

Introduction: Conjugated linoleic acid (CLA), a ruminant-derived fatty acid abundant in dairy and red meat, has been consumed as a dietary supplement for its purported health benefits. We aimed to prospectively investigate the association of CLA intake with coronary heart disease (CHD) risk in U.S. individuals and to explore the synergistic roles of host metabolome and gut microbiome. Methods: We followed 64,090 women from the Nurses’ Health Study (1986–2018) and 42,659 men from the Health Professionals Follow-Up Study (1986–2016) for incident CHD. Intake of CLA was assessed using 7-day diet records in the Lifestyle Validation Study (LVS) and validated food frequency questionnaires administrated quadrennially in the cohorts. Plasma metabolomic signatures responsive to CLA were derived using elastic net regression in the LVS (n=1,143). The gut microbiome was profiled in the Men’s LVS (n=307) and in a CHD case–control substudy (n=152). Results: Dietary CLA intake was associated with distinct alterations in the plasma metabolome and gut microbiome. Higher CLA intake was linked to elevated phospholipids, primarily phosphatidylethanolamine (PE 34:0) and phosphatidylcholine (PC 34:4, 36:4), and triacylglycerols with 0-2 double bonds (TAGs 49:2, 50:0, 50:1, 51:1), all positively linked to CHD risk. Conversely, CLA intake was related to lower levels of TAGs with more double bonds (54:3, 54:4). Higher CLA intake was associated with reduced Bacteroides clarus abundance (FDR = 0.03), which was inversely linked to ruminant fat–derived TAG 51:1. In the case–control substudy, higher B. clarus abundance predicted lower CHD risk ( P < 0.25). A multi-metabolite signature for cis-9, trans-11 CLA, composing 86 metabolites, showed a strong correlation with dietary intake (Spearman r = 0.73) and predicted higher fatal CHD risk (HR 1.82, 95% CI 1.11–2.99). In contrast, the metabolomic signature for trans-10, cis-12 CLA showed no significant association. In the two cohorts, higher cis-9, trans-11 CLA intake was also consistently associated with a greater risk of CHD (pooled HR for highest vs lowest quintile: 1.27, 95% CI 1.11–1.45), driven by fatal CHD (HR 1.50, 95% CI 1.24–1.81). Conclusions: Higher intake of conjugated linoleic acid, particularly cis-9, trans-11 isomer, was associated with increased risk of fatal CHD and unfavorable plasma metabolomic profiles. These associations were potentially modulated by specific gut microbial taxa, such as Bacteroides clarus .

Correction for Wang et al., Mechanistic insights into dengue virus inhibition by a clinical trial compound NITD-688

Prevalence and genotypic distribution of non-epidermolytic ichthyosis in Italian Golden Retrievers

Non-epidermolytic ichthyosis (NEI) is a hereditary skin disorder affecting several dog breeds, most notably the Golden Retriever. It is primarily caused by a loss-of-function variant in the PNPLA1 gene, while a second, less common form is associated with a deletion in the ABHD5 gene. This retrospective study aimed to assess the prevalence and temporal trends of both mutations in Golden Retrievers tested in Italy between 2017 and September 2025. A total of 508 genetic tests were analyzed, including 463 dogs tested for the PNPLA1 mutation, 42 for the ABHD5 deletion, and 3 for both variants. DNA was extracted from blood or buccal samples and analyzed by real-time PCR followed by confirmatory Sanger sequencing. Among the PNPLA1 tested dogs, 42% were clears (wt/wt), 37% carriers (wt/mut), and 21% affected (mut/mut), with calculated allele frequencies of 60% wild-type and 40% mutant. A significant temporal decline in mutant allele frequency was observed, accompanied by an increasing number of animals tested over time, suggesting growing interest in genetic screening and its impact on selective breeding. Conversely, all dogs tested for the ABHD5 deletion were wild-type, supporting its rarity in the breed. Overall, these findings confirm that PNPLA1 -related ichthyosis remains one of the most prevalent hereditary disorders in Golden Retrievers, although its frequency is decreasing. The results emphasize the effectiveness of genetic testing in disease prevention and highlight the importance of continued monitoring to maintain genetic health within the breed.

Abstract WE457: Comparative Analysis of Hypertension Management Plans Generated by Large Language Models

Background: The use of large language models (LLMs) in clinical practice, medical education, and by patients is increasing rapidly. It is essential to ensure that information provided by these artificial intelligence (AI) chatbots is accurate and safe. Our goal was to analyze and compare hypertension treatment plans generated by popular LLMs and identify their strengths and limitations. Methods: ChatGPT-4o, Claude, ClinicalKey AI, Copilot (Microsoft), DeepSeek-V3, Dyna AI, Google Gemini, Grok, Meta AI, OpenEvidence, Perplexity, and Pi were prompted to generate a treatment plan for stage two hypertension. Five blinded reviewers scored each response in three domains: adherence to clinical guidelines, detail/clarity, and reliability/safety (sources provided/emphasis on seeing a healthcare professional). Mean scores for each domain were calculated and summed for a composite score. The responses were also analyzed qualitatively by three reviewers. Results: Perplexity received the highest composite score (8.2 out of 9), followed by OpenEvidence (7.7 out of 9). Dyna AI had the lowest overall score (3.7 out of 9), followed by Pi (4.8 out of 9), ClinicalKey AI (4.9 out of 9), and Meta AI (4.9 out of 9). Perplexity (3 out of 3), Grok (2.8 out of 3), and OpenEvidence (2.7 out of 3) had the highest scores for detailed and clear responses, while DynaAI had the lowest for both detail/clarity (1 out of 3) and reliability/safety (1 out of 3). ChatGPT-4o had the highest score for adherence to guidelines (2.7 out of 3) while Pi had the lowest (1.5 out of 3). Analysis of Variance (ANOVA) statistical test showed statistically significant differences across every subscore domain and composite scores (p=0.00125 for adherence to guidelines, p<0.00001 for detail/clarity, p=0.00003 for sources/seeing a professional, p<0.00001 for composite scores). Qualitatively, the LLMs tended to adhere to guidelines and provide sufficiently detailed management plans, but often did not provide sources and/or advise users to see a healthcare professional. Conclusions: The LLMs nearly always provided detailed management plans for stage two hypertension that adhere to clinical guidelines. However, there was significant variability in quality by different chatbots. Notably, medicine-specific LLMs were not necessarily superior to LLMs used by the general public. AI chatbots may require greater regulation to ensure that they inform users to see a healthcare professional and provide reliable sources.

Abstract MPWE26: Breastfeeding and Early Childhood Cardiovascular Health in the ECHO Program

Background: Breastfeeding is linked with single indicators of maternal and child cardiovascular health (CVH), but its association with comprehensive childhood CVH as defined by AHA Life’s Essential 8 (LE8) criteria is unclear. We investigated whether breastfeeding duration and exclusivity were associated with early childhood CVH. Methods: We analyzed data from mother-child dyads in the Environmental influences on Child Health Outcomes (ECHO) Cohort and with data on breastfeeding and child CVH outcomes at age 3-8 years. Breastfeeding duration was categorized as never/<1 month [reference], 1 to <6 months, 6 to <12 months, and ≥12 months, and secondarily, exclusive breastfeeding duration as never [reference], <3 months, 3 to <6 months, and ≥6 months. The outcome was a modified LE8 using available BMI, blood pressure, diet, sleep, and secondhand smoke data and analyzed as a continuous composite score and dichotomized as high CVH (LE8 ≥80) vs moderate/ low CVH (LE8<80). Modified Poisson regression models were used, with multiple imputation for missing covariates and adjustment for maternal age at birth, parity, education, employment, pre-pregnancy BMI, hypertensive disorders of pregnancy or gestational diabetes, cesarean delivery, and gestational age at birth. Results: The cohort included 5,278 maternal-child dyads (median child age 5.7 years [IQR:4.7-6.9]); 38.5% never/ breastfed for <1 month, 16.4% 1 to <6 months, 25.1% 6 to <12 months and 20.0% for ≥12 months; and 29.9% exclusively breastfed for <3 months, 16.8% 3 to <6 months, and 11.7% ≥6 months. Mean LE8 score was 82.5 (SD:14.5), and 62.4% had high CVH. Compared to never/<1 month of breastfeeding, children breastfed for 6 to <12 months (mean diff: 2.16; 95% CI:1.00–3.32) and ≥12 months (mean diff: 3.66; 95% CI:2.26–5.07) had higher LE8 scores and greater likelihood of high CVH (aRR:1.21; 95% CI:1.06–1.39 and aRR:1.31; 95% CI:1.13–1.52, respectively) ( TABLE 1 ). Exclusive breastfeeding for 3 to <6 months and ≥6 months was associated with higher LE8 scores (mean diff: 1.97; 95% CI:0.66–3.29 and mean diff: 3.36; 95% CI:1.85–4.87, respectively), and increased likelihood of high CVH (aRR:1.18; 95% CI:1.01–1.37 and aRR:1.30; 95% CI:1.11–1.52, respectively) ( TABLE 2 ). Conclusions: Longer duration of any or exclusive breastfeeding was associated with higher LE8 scores in early childhood. These results from across the US underscore breastfeeding as a modifiable early-life exposure with potential CVH benefits.

Defective RNA Polymerase III sensing of mitochondrial DNA in pulmonary epithelial cells impairs type I IFN immunity to SARS-CoV-2

The clinical spectrum of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection ranges from asymptomatic cases to critical COVID-19 pneumonia. To investigate the role of host genetics in susceptibility to critical COVID-19 and identify pathophysiological mechanisms and pathways, we analyzed whole-exome and whole-genome sequencing data from the COVID Human Genetic Effort. We identified 10 rare, monoallelic predicted loss-of-function variants in 18 patients in POLR3A and POLR3C encoding two subunits of RNA polymerase III (POL III), a nuclear multisubunit enzyme, which has been implicated in cytosolic DNA sensing. These variants were deleterious for expression of full-length POLR3A and POLR3C proteins. We demonstrate that human pulmonary A549-hACE2 cells with reduced POLR3A or POLR3C expression exhibit impaired type I IFN responses to transfected mitochondrial DNA (mtDNA) or SARS-CoV-2 infection, together with increased viral replication. Mechanistically, we show that SARS-CoV-2 induces cellular mtDNA release via oligomerization of the mitochondrial voltage-dependent anion channel under virus-induced oxidative stress, enabling POL III–mtDNA interaction. These findings establish POL III as a sensor of endogenous mtDNA released during viral infection and indicate that autosomal dominant POL III haploinsufficiency may predispose individuals to critical COVID-19.

Dynamic optimization of high-bandwidth multi-receiver signals for civil aircraft flight tests telemetry

Civil aircraft flight tests are distinguished by parallel multi-task operations, multi-sensor collaboration, and high-frequency data acquisition. The exponential growth in telemetry data generated in a single flight imposes stringent requirements on the high-bandwidth transmission capabilities of telemetry systems and the real-time optimization of multi-receiver signals. This study focuses on addressing the dynamic optimization of high-bandwidth multi-receiver signals during flight tests, aiming to enhance demodulation accuracy and ensure stable real-time transmission of large-scale telemetry data. A channel model for complex scenarios was constructed to analyze signal redundancy and dynamic switching in multi-receiver links, while an improved routing protocol integrated with multi-dimensional signal evaluation was investigated. A modular simulation model for the multi-receiver telemetry system was developed within a highly dynamic wireless ad hoc network environment. By enhancing the AODV and LEACH routing protocols, multi-path backup and energy consumption optimization were achieved. A multi-dimensional signal optimization algorithm was utilized to dynamically evaluate and fuse bit synchronization and frame synchronization via voting, enabling real-time selection of the optimal received signal in complex environments. The findings indicate that high-bandwidth multi-receiver signal optimization and enhanced routing strategies effectively alleviate challenges associated with multipath fading, channel obstruction, and high-speed mobility in flight test scenarios, thereby facilitating real-time channel monitoring and dynamic switching while improving data timeliness and accuracy. Simulation results demonstrate that under high-dynamic conditions, the proposed algorithm reduces the average BER to 3e -6 at 10 dB SNR, improves the frame synchronization success rate to 98.9%, and reduces invalid channel switching frequency by approximately 84% compared to traditional methods.

Abstract TH800: Shared Clinical and Genetic Determinants of Cardiometabolic Disease and Lung Cancer

Introduction: Cardiometabolic disease (CMD) may influence lung cancer risk through shared biological pathways and treatment effects. We examined clinical and genetic factors underlying comorbidity between CMD and lung cancer. Method: We analyzed 214,096 adults with CMD or lung cancer using electronic health record data. Variables included age, sex, body mass index, smoking status, CMD subtype, statin use, and cancer history. We used logistic regression to estimate the association of CMD phenotypes with lung cancer incidence, adjusting for demographics and smoking. We conducted age-stratified analyses and performed Mendelian randomization using genome-wide association summary statistics to assess genetic correlations between CMD traits and lung cancer. Results: Atrial fibrillation (OR 1.24, 95% CI 1.19–1.29), myocardial infarction (OR 1.20, 95% CI 1.16–1.25), heart failure (OR 1.35, 95% CI 1.30–1.39), and coronary atherosclerosis (OR 1.36, 95% CI 1.31–1.41) were associated with higher lung cancer risk. Dyslipidemia (OR 0.75, 95% CI 0.66–0.86), hypertension (OR 0.69, 95% CI 0.65–0.73), and type 2 diabetes (OR 0.95, 95% CI 0.92–0.99) were inversely associated. Statin use modestly increased risk (OR 1.05, 95% CI 1.01–1.08), with fluvastatin and atorvastatin showing the strongest effects. Mendelian randomization indicated a genetic association between type 2 diabetes liability and lung cancer (OR 1.14, 95% CI 1.08–1.21). Conclusion: We found that cardiometabolic conditions and their treatments are associated with lung cancer risk through both clinical and genetic pathways. Understanding these shared determinants may improve risk stratification and support integrated prevention strategies targeting cardiovascular and pulmonary health.

Abstract MPWE29: Maternal Cardiovascular Disease Events Within Four Years of Delivery Following Pregnancy Complicated by COVID-19 Infection in 2020-2022.

Introduction: COVID-19 has been associated with increased risk during pregnancy and increased long-term cardiovascular disease (CVD) risk; however, few studies have examined CVD risk following pregnancy complicated by COVID-19. Therefore, our study assessed the pre-pregnancy and trimester-specific impact of COVID-19 on incident maternal CVD within four years of delivery in South Carolina (SC). Methods: We utilized SC vital records linked to hospitalization/emergency department records, and SC health department data on COVID-19 infections. COVID-19 exposure was defined as none, pre-pregnancy, during the 1 st or 2 nd trimester, or during the 3 rd trimester. CVD events were defined broadly based on ICD-10 codes: B33, G45.8, G45.9, I11, I13.0, I13.2, I20-I21, I24-I26, I30, I32, I40-I42, I44-I50, I51.4, I60-I67, I70-I72, I74, I75, I82, O87.1, O87.3, O90.3, R00.0, R00.1, R57.0. Women were followed from 7 days after delivery through 2023. Pregnancies were matched (on race/ethnicity, delivery quarter time, Medicaid eligibility, and maternal age), adjusted (for education, rural residence, smoking, firstborn, previous preterm delivery, and pre-pregnancy BMI, hypertension, and diabetes) and analyzed using a modified Poisson model. Cox proportional hazard models were also used in unmatched adjusted analyses. Results: Of 121,510 singleton deliveries to women aged 12-52 years f rom March 2020 through December 2022, 8.0% had a COVID-19 infection pre-pregnancy, 6.1% in the 1 st or 2 nd trimester, and 3.8% in the 3 rd trimester, and 3.0% had a CVD event. In matched analyses, COVID-19 infection during the 1 st and/or 2 nd trimester [risk ratio (RR)=1.24, 95% CI: 1.05-1.45] increased the risk of CVD events compared to pregnancies without COVID-19 after adjustment. In contrast, COVID-19 infection pre-pregnancy (RR=0.93, 95% CI: 0.78-1.10) or during the 3 rd trimester (RR=0.93, 95% CI: 0.76-1.15) was not associated with increased risk of CVD. Cox proportional hazard model results were similar in unmatched adjusted analyses for COVID-19 infection during the 1 st or 2 nd trimester (hazard ratio (HR)=1.27, 95% CI: 1.11, 1.46), pre-pregnancy (HR=0.97, 95% CI: 0.83, 1.14) and during the 3 rd trimester (HR=0.95, 95% CI: 0.79, 1.13). Conclusion: Risk of CVD was elevated for pregnant women with COVID-19 infection during the 1 st or 2 nd trimester but not during the 3 rd trimester or pre-pregnancy, emphasizing the importance of preventive measures like vaccination to curb potential long-term risk.

Mechanics of dust bathing in birds

Dust bathing is a widespread behavior among birds that helps remove ectoparasites, but its mechanical basis remains poorly understood. Here we show that wing flapping during dust bathing generates sand-particle collisions sufficient to dislodge feather mites. Using high-speed video analysis across 34 bird species, we found that wing-flapping frequency scales inversely with body mass ( f ∼ M - 0.20 ), ranging from 1 to 10 Hz. Since feather motion matches wing motion, we use mechanical shakers to simulate dust-bathing dynamics. Mite-removal experiments reveal that parasite dislodgment efficiency increases rapidly up to around 5 Hz and plateaued thereafter. The frequencies around 5 Hz are sufficient to achieve maximal cleaning efficacy. Collision modeling indicates that the impact force from sand particles exceeds mite adhesion forces, thus collision is the dominant removal mechanism. Mites are removed within seconds when sand is present, suggesting that sand plays an active mechanical role in parasite dislodgment. These findings not only elucidate a physical mechanism underlying dust bathing, but also suggest that birds may have evolutionarily tuned their flapping behavior to optimize parasite removal through substrate interaction.

An integrated model of financial socialization, technology, and financial capability in predicting financial well-being

This study develops and empirically tests an integrated framework that explains how financial socialisation, technological factors, and financial capability jointly shape financial behaviour and in an emerging economy context. Using data from 306 Vietnamese adults, the study applies Partial Least Squares Structural Equation Modelling to assess direct, mediating, and moderating effects. The results show that both family financial socialization and artificial intelligence significantly enhance financial behaviour and financial well-being, with financial behaviour mediating these relationships. Artificial intelligence exerts a stronger influence on financial behaviour than family financial socialisation, while its impact on financial well-being operates primarily through behavioural pathways. Financial literacy and digital trust significantly strengthen the effect of artificial intelligence on financial behaviour, although the moderating effects are relatively modest. Financial well-being is positioned as the ultimate outcome of the model, and the findings confirm that improvements in well-being are largely driven by behavioural adjustments rather than direct technological exposure alone. The study offers theoretical contributions by integrating social, technological, and capability-based elements into a unified financial well-being framework and highlights the conditional roles of digital trust and financial literacy in shaping AI-driven financial behaviour. It also provides practical implications for financial education and responsible digital finance adoption to enhance financial resilience and long-term well-being.

Abstract WE546: Sex-Specific Lipidomic Signatures Associated With Obstructive Sleep Apnea Severity

Background: Obstructive sleep apnea (OSA) is a prevalent cardiometabolic disorder with known sex differences in presentation and diagnosis. The apnea–hypopnea index (AHI) often underrepresents disease severity in females. Although lipidomic alterations have been reported, few studies have systematically stratified these associations by sex or BMI. Hypothesis: We hypothesized that AHI and related OSA traits are associated with distinct circulating lipid profiles, and that these associations differ by sex. Methods: We analyzed data from the Human Phenotype Project, which includes home sleep apnea testing and targeted lipidomic profiling. A total of 143 lipid species were analyzed.Multivariable regression models were adjusted for age, sex, and BMI, with additional sex-stratified and interaction analyses performed. Statistical significance was defined as a false discovery rate <0.05. OSA was defined as an AHI ≥15 events/hour. Results: Among 2887 adults (51.6 % female), 25.1% had OSA. In this subgroup, the mean age was 54.54 ± 7.75 years, mean BMI was 28.67 ± 4.11 kg/m 2 , mean AHI was 24.65 ± 9.90 events/hour. Among these participants, 32.8% had hypertension, 1.3% had diabetes, and 24.5 % reported excessive daytime sleepiness. ( Table 1). In BMI-adjusted models, higher AHI was associated with elevated triacylglycerols (TAGs). For example, TAG (47.0) showed β = 0.007 (95% CI 0.002 - 0.011, FDR = 0.014), and TAG (51.1) showed β = 0.009 (95% CI 0.005–0.013, FDR = 0.001). Sex- stratified analysis identified 14 statistically significant lipids in female – 13 TAGs (TAG 47.0, 47.1, 49.1, 50.0, 50.5, 51.1, 51.3, 51.4, 53.4, 54.1, 56.3, 58.2, 60.4) and one Phosphatidylserines (PS_37.2) that remained significant after BMI adjustment. No significant associations were found in males (Figure 1) . Excessive daytime sleepiness was inversely associated with ceramides such as Cer d18:1/24:1 (β = –0.120, 95% CI –0.188 to –0.052, FDR = 0.031). Interaction testing confirmed pronounced sex differences in BMI-adjusted analysis, with all statistically significant associations having sex interaction p-value<0.05. Conclusions: OSA traits, particularly AHI, are linked to distinct lipidomic signatures in females. A consistent set of 14 female-driven lipids—predominantly TAGs—persisted after BMI adjustment, suggesting BMI-independent pathways. These robust, female-specific TAG signals may represent reliable biomarkers of cardiometabolic risk in women with OSA.

Abstract WE541: Early Detection of Subclinical Myocardial Dysfunction in Diabetes Using Leak-Safe Machine Learning on Routine Clinical Data

Background: Diabetic patients may develop subclinical myocardial dysfunction despite preserved ejection fraction (EF≥50%) and normal Doppler indices. Speckle-tracking echocardiography (STE) detects these abnormalities earlier, but STE is resource-intensive and not universally available. A machine-learning model using routine clinical data could identify high-risk patients for targeted imaging and earlier intervention. Objective: To develop and validate a leakage-safe machine-learning approach identifying subclinical myocardial dysfunction—defined as abnormal STE despite EF≥50% and normal Doppler parameters—in adults with diabetes. Methods: We analyzed a single-center cohort of diabetic adults. Positive class: EF≥50%, normal Doppler indices (E/e′≤14, tricuspid regurgitation velocity≤2.8 m/s, left atrial volume index≤34 mL/m^2), and abnormal global longitudinal strain (GLS). Negative class: same EF/Doppler criteria with normal GLS; others were excluded. Input features included demographics, vitals, anthropometrics, comorbidities, medications, and laboratories. All STE variables were excluded to prevent circularity. Preprocessing (median imputation, one-hot encoding, scaling) used a unified pipeline. Grouped 5-fold cross-validation by patient ID prevented patient-level leakage. Primary metrics: AUROC and average precision (AP). Models tested: logistic regression, random forest, and XGBoost. Results: Among 233 eligible patients, 199 (85.4%) were GLS-abnormal and 34 (14.6%) GLS-normal. XGBoost achieved perfect discrimination (AUROC 1.000, AP 1.000) in cross-validation and out-of-fold testing. Random Forest performed strongly (CV AUROC 0.963±0.022, AP 0.994±0.003; OOF AUROC 0.962, AP 0.994), surpassing logistic regression (CV AUROC 0.788±0.059, AP 0.961±0.008; OOF AUROC 0.800, AP 0.961). Routine clinical variables accurately identified STE-positive patients. Given exceptional XGBoost performance, sensitivity analyses and external validation are warranted. Conclusions: Leakage-aware ML using standard clinical data can flag diabetic patients with STE-defined subclinical dysfunction despite preserved EF and normal Doppler indices, enabling earlier risk stratification and targeted imaging before symptomatic heart failure. Future work includes multi-center validation, calibration assessment, and clinical workflow integration.

Lethal toxin–equipped effector cells for the potential treatment of cancer

Lethal toxins could become potent therapies against cancer, but their clinical utility is limited by adverse events upon systemic administration. These could be reduced if the toxins were delivered by effector cells that specifically infiltrate cancers, thereby releasing toxins locally into the tumor microenvironment. One of the challenges underlying this strategy is that cells delivering toxins would have to be resistant to them. We address this obstacle by showing that effectors derived from transformed human cell lines genetically engineered for resistance to bacterial adenosine diphosphate ribosylating toxins (ADPRTs), including Pseudomonas aeruginosa exotoxin A (PE), can produce targeted immunotoxins that specifically kill cancer cells expressing cognate tumor-associated antigens. Resistance to immunotoxins was achieved by knockout of genes in the diphthamide biosynthesis pathway ( DPH1-4 ) required for the posttranslational modification of eukaryotic elongation factor 2 (EEF2) that is the target of ADPRTs or by mutation of EEF2 itself. We show that engineering resistance to ADPRTs, one of the most potent toxins acting on human cells, is essential to achieve robust function of armored effector cell lines. This work establishes a critical step on the path to equip effector cells with the ability to deliver powerful toxins to cancer cells and introduces a platform to investigate extension to primary autologous or allogeneic therapeutic cell types.

HMGB1 reduce DNA damage by binding KU70 to activate NHEJ pathway in colorectal cancer cells after radiation

DNA damage-induced by radiotherapy is a critical factor in promoting the death of colorectal cancer cells (CRC). Although high mobility group box 1 (HMGB1) reportedly plays a vital role in tumor radioresistance by modulating DNA damage repair, the precise mechanisms remain unclear. In this study, HMGB1 knockdown markedly enhanced cell apoptosis after radiation. HMGB1 downregulation significantly inhibited DNA damage repair and reactive oxygen species (ROS)-mediated redox homeostasis after irradiation in CRC cells. Mechanistically, HMGB1 interacts with KU70 via its region spanning residues 95–163. This interaction subsequently activates the non-homologous end joining (NHEJ) pathway to facilitate DNA damage repair, ultimately leading to reduced radiation-induced cell apoptosis. KU70 silencing showed the same effect as HMGB1 depletion mediated cell apoptosis and DNA damage response both in vitro and in vivo . Additionally, HMGB1 and KU70 were overexpressed in CRC tissues. Analysis of the GEPIA database indicated that elevated levels of both genes showed a trend toward association with poor patient prognosis, although this did not reach statistical significance. The current study revealed that HMGB1 may promote DNA damage repair through KU70 and its mediated NHEJ pathway to affect apoptosis in CRC cells after irradiation. Thus, targeting the HMGB1/KU70/NHEJ axis may be a potential therapeutic target to promote the response of CRC to radiotherapy and in-depth study of the specific mechanism of this axis in CRC radioresistance will help to the develop more effective treatment strategies.

Abstract TH878: Explainable OCT-Based Atherosclerotic Plaque Segmentation Using Convolutional Neural Networks and Large Language Models

Objective: To develop and evaluate an explainable framework for automated segmentation of atherosclerotic plaque morphological features in optical coherence tomography (OCT) images, integrating convolutional neural networks (CNNs) for image analysis and large language models (LLMs) for natural-language explainability. Background: Cardiovascular disease remains the leading cause of mortality worldwide, with atherosclerosis at its core. High-resolution imaging enables detailed visualization of plaque structures such as the lumen, fibrous cap, lipid core, and vasa vasorum. However, manual interpretation of OCT slices is time-consuming and requires expert knowledge. While CNN-based segmentation models can automate feature identification, they often lack interpretability. Existing explainability approaches, such as heat maps, highlight image regions used for prediction but fail to provide semantic reasoning useful to clinicians and researchers. This study explores the combined use of CNNs and LLMs (GPT, Claude, Gemini) to generate interpretable, human-readable explanations accompanying automated segmentation results. Methods: We utilized a curated OCT dataset of 25,698 annotated slices from 103 patients across two cardiovascular research centers. Each slice included manual annotations of key plaque morphological features (lumen, fibrous cap, lipid core, and vasa vasorum). A CNN-based segmentation model was trained using a 5-fold cross-validation design to ensure robust evaluation. LLMs were incorporated post hoc to interpret the CNN’s predictions, explaining morphological feature identification and potential clinical implications in natural language. The semantic quality, factual accuracy, and consistency of generated explanations across different LLMs were qualitatively assessed. Results: Quantitative and qualitative evaluation of the CNN segmentation performance and LLM-generated explanations is currently underway. Results, including segmentation accuracy, comparative model performance, and semantic analysis of LLM explanations, will be presented at the conference. Conclusions: This study introduces an explainable AI framework that unites CNN-based image segmentation with language-based reasoning for OCT plaque analysis. By complementing automated plaque identification with clear, contextually grounded explanations, this approach aims to enhance interpretability, reduce cognitive load for clinicians, and promote transparency in cardiovascular imaging research.

Abstract MPTU18: Single Family Zoning and Optimal Cardiovascular Health Measured using the American Heart Association’s Life’s Essential 8 in the Multi-Ethnic Study of Atherosclerosis

Background: Single-family zoning (SFZ) is an urban planning regulation that prevents neighborhoods zoned for single-family homes from developing multifamily housing units, which are often more affordable and disproportionately utilized by lower-income individuals. We evaluated whether SFZ was related to cardiovascular health. Methods: We used 2010-2011 data from the Multi-Ethnic Study of Atherosclerosis (MESA) study and 2010-2012 zoning data of the 6 MESA study sites. SFZ was defined as the percentage of land zoned as single-family within a one-mile radius of the study participants’ homes. Cardiovascular health was defined as the American Heart Association’s Life’s Essential Eight (LE8) composite measure, comprised of diet, physical activity, smoking status, body mass index, cholesterol level, fasting blood glucose, and blood pressure. Higher scores implied better health. We also calculated a measure without sleep (LE8-no sleep) since sleep data were missing on 50% of participants. Multivariable linear mixed models, controlling for sex, education, income, and age, and accounting for clustering by census tract, were used to examine the relationship between SFZ and LE8/LE8-no sleep, overall and by MESA site. Results: Among MESA participants (N=3,550), SFZ was greatest in buffers around those residing in Winston-Salem County, NC (median: 70%) and lowest in Upper Manhattan/Bronx, NY (0%). Overall, living in an area with 10% more SFZ was associated with higher LE8-no sleep scores (β: 0.21, 95% CI: 0.005, 0.42). In stratified analyses, a 10% increase in SFZ was associated with greater LE8-no sleep in Baltimore County (β: 1.14, 95% CI: 0.47, 1.81) and Los Angeles County (β: 0.80, 95% CI: 0.14, 1.45). In Chicago, a 10% increase in SFZ was associated with lower LE8-no sleep (β: -1.42, 95% CI: -2.14, -0.70) and LE8 (β: -1.20, 95% CI: -2.11, -0.29). Conclusion: The relationship between SFZ and cardiovascular health may differ between locations. As cities consider creating or rescinding policies of exclusionary zoning, the implications on health should be considered.

Condensate-mediated shape transformations of cellular membranes by capillary forces

Phase-separated biomolecular condensates with liquid-like properties play a key role in the organization and compartmentalization of the intracellular environment. Condensate-mediated capillary forces acting on membranes drive physiologically important reshaping of membrane-bound organelles, such as vacuoles and autophagosomes. Here, we explore condensate-mediated membrane shape transformations. We employ in planta live-cell imaging, an in vitro reconstitution system with tunable interfacial tension, and computer simulations of an elastic membrane model to describe three morphologies of membrane structures localized at condensate interfaces: tubes, sheets, and cups. We find that the forces associated with high interfacial tension drive the formation of stable sheets, while tubes and cups prevail at lower interfacial tension. We calculate the free energies of each membrane shape and identify the energy barriers that govern the transitions between the shapes. With this approach, we find that shape transformations depend on the history of the interfacial membrane and exhibit a tube-to-cup hysteresis. These findings indicate that temporal control of condensate surface properties can mediate the morphogenesis of cup-like structures in cells, such as the formation of “bulbs” within plant vacuoles. Our results further generalize how the interplay of condensates and membranes contributes to intracellular organization.