eBioMedicine

Reduced nicotinamide mononucleotide (NMNH) is a reduced NAD precursor with reported NAD- augmenting activity in preclinical models; however, controlled human data remain limited. This was a randomized, double-blind, placebo-controlled, parallel-group phase I trial evaluating oral NMNH-Ca in healthy adults aged 40-65 years. Eighty participants received placebo or NMNH-Ca 125, 250, or 500 mg once daily for 90 days. The primary objective was safety and tolerability. Whole-blood NAD was assessed as the key pharmacodynamic endpoint, including a 24-hour post-dose substudy, with biomarker-derived blood phenotypic age, treadmill-based six-minute walk distance, body mass index, and SF-36 domains analyzed as exploratory outcomes. NMNH-Ca was well tolerated at all doses, with no serious adverse events, treatment-related adverse events, or discontinuations. In the acute substudy, whole-blood NAD increased after single-dose NMNH-Ca, with peak mean concentrations at 12 hours. Over 90 days, NAD increased in a dose-related pattern; Day 90 mean changes from baseline were 2.33 {+/-} 18.53 M with placebo and 8.22 {+/-} 10.25, 15.85 {+/-} 11.16, and 39.90 {+/-} 14.11 M with NMNH-Ca 125, 250, and 500 mg, respectively. Exploratory analyses showed hypothesis-generating favorable signals in blood phenotypic age, treadmill-based six-minute walk distance, and health-related quality of life, most consistently at 500 mg. Oral NMNH-Ca was safe and pharmacodynamically active over 90 days, supporting larger and longer confirmatory trials with prespecified geroscience endpoints and tissue-relevant NAD metabolomics.

Self-repairs, in-utterance revisions in which a speaker abandons and reformulates their speech, are a promising interpretable marker for speech-based cognitive screening. Detecting them automatically is difficult because a self-repair is defined by its relationship to surrounding speech rather than by fixed lexical cues. On the DementiaBank ADReSS corpus, we compared the capability of generative LLMs under a five-condition prompt ablation against a fine-tuned DistilBERT token classifier at detecting self-repairs. GPT-5 performed best (test F1 = 0.73) and was largely insensitive to prompt design, whereas the LLaMA (open-weight alternative) was both weaker and far more prompt-sensitive (test F1 = 0.47). DistilBERT, nearly 100 times smaller, matched the open-weight LLM at a fraction of the computational cost. These results suggest that a locally deployable encoder, given sufficient in-domain annotation, is a more plausible route to clinical self-repair detection than scaling model size or prompt complexity.

Tuberculosis (TB) remains a major public health challenge, exerting profound socio-economic burdens and causing debilitating illness in approximately 2.5 million individuals across Africa annually. Optimized large-scale treatment regimens, such as NAT2-genotype adjusted dosing, could improve patient outcomes and strengthen healthcare systems. However, fully addressing the complexity of multi-drug TB treatment responses requires consideration of the entire pharmacogenomic (PGx) landscape, particularly within African populations, which are both genetically diverse and critically understudied. In this study, we predict NAT2 genotypes and phenotypes in specific African populations, and we extend TB PGx research beyond well-established biomarkers. Current bioinformatic prediction tools were used to evaluate individual- and population-specific variation in genotype and next-generation sequencing data from 2,143 individuals across 20 African population groups, spanning ten PGx genes associated with multi-drug TB treatment and response. Most predicted functionally deleterious variants occurred at low frequencies (MAF < 0.01) and were observed in only one of the 20 populations. The Khomani and Nama populations had a distinctly higher proportion of NAT2 fast metabolizer phenotypes than other African populations, indicating a lower risk of INH overexposure and possibly different dosage requirements in these groups. These findings highlight both the potential and current limitations of functional prediction for absorption, distribution, metabolism and excretion (ADME) variants, and the transferability of their predictive value between African population groups. With the increasing accessibility of next-generation sequencing, alongside the development of comprehensive databases capturing African variation and advances in computational algorithms, the cumulative impact of genetic variation on TB drug response can be more accurately captured, thereby informing precision treatment strategies.

Background: ERAS protocols reduce hospital stay by 1.88 days and complications by 29% globally, but their one-size-fits-all paradigm, validated at sea level, may fail at high altitude where chronic hypoxia and population-specific genetic adaptations remodel baseline physiology. No study has quantified ERAS effect weight shifts at high altitude or proposed a theoretical model to explain the gap. Objectives: To evaluate three dimensions of plateau ERAS remodeling: (i) risk factor weight shift, (ii) traditional marker failure, (iii) genetic background modification, and propose the PAERS (Plateau Adaptation-ERAS Remodeling Syndrome) risk stratification model tailored to altitude. Methods: Retrospective cohort of 612 adults undergoing elective laparoscopic cholecystectomy (2018-2023) at Qinghai Red Cross Hospital (2260 m). Three analytical tiers: (1) multivariable regression comparing risk factor coefficients against plain-altitude benchmarks; (2) restricted cubic spline and interaction modeling for Hb, SpO2, and LOS; (3) inferential genetic modifier analysis using population-level EPAS1 carrier rates. Primary outcomes: LOS and complication rate. Results: Three-dimensional shift was observed: (1) Weight Remodeling: BMI replaced sex as primary risk factor (OR = 1.86, P < .001), surgeon variability amplified (F = 6.33 vs plain benchmark 2-4, an ~58% increase in F-statistic ratio, P < .001); (2) Marker Failure: Hb showed J-type relationship with LOS (Hb x SpO2 interaction beta = -0.0095, P = .009), with effect reversal across SpO2 strata (Plateau Hemoglobin Paradox); (3) Genetic Modification (population-level inference): ~70% EPAS1 carrier rate (range 57-85% across studies) suggests HIF-2alpha pathway is a baseline modifier that must be accounted for. Three falsifiable predictions were proposed. Conclusions: High-altitude ERAS faces three challenges: effect weight remodeling, biomarker failure, and genetic background calibration. The PAERS hypothesis proposes an integrated risk stratification model, shifting from one-size-fits-all to altitude-aware, patient-specific protocols.

Biological aging occurs heterogeneously across individuals and organs. However, current measures of biological age incompletely capture organ-specific differences in health and disease risk. Because chest CT visualizes multiple thoracic organs, it offers an opportunity to quantify structural aging across organ systems. Here, we developed MOSAIC-Age, a framework characterizing eight organ-specific aging clocks on chest CT. The clocks were developed and validated using 9,971 CT scans from CT-RATE and MIDRC, and subsequently locked and applied to two independent prospective cohorts with 35,293 participants from the National Lung Screening Trial and Genetic Epidemiology of COPD study. CT-derived biological age gaps (BAGs) were examined in relation to lifestyle and socioeconomic factors, prevalent comorbidities, incident chronic diseases, and all-cause and cause-specific mortality. Higher BAGs, indicating organs that appeared older on CT than expected for their chronological age, were broadly associated with adverse health characteristics, chronic disease burden, and increased mortality risk. Multiple disease outcomes were associated with aging across several organs, whereas in multivariable analyses including all eight organ-specific BAGs, the remaining associations were more organ specific. A greater number of markedly older-appearing organs and a faster pace of aging were each associated with higher mortality. Together, these findings demonstrate that routine chest CT captures both shared and organ-specific patterns of biological aging and establish CT-derived organ aging as a quantitative imaging biomarker for assessing multi-organ health and long-term disease risk.

Purpose: Whether host germline genetic variation determines tumor drug response remains underexplored. We evaluated whether EPAS1 (HIF-2) adaptive loss-of-function variants, enriched in high-altitude-adapted populations, predispose HCC to primary antiangiogenic TKI resistance through a HIF-2/STC2 signaling axis. Experimental Design: We integrated five independent data sources: the QHRCH-HCC retrospective cohort (n = 1,396), multi-ancestry iPSC-derived endothelial cell transcriptome data (GSE160906), TCGA pan-cancer data (LIHC, KIRC, LUAD, BRCA), GDSC2 pharmacogenomics (n = 951 cell lines; 11 antiangiogenic TKIs), and DepMap dependency data. The AESI_score integrated altitude, AFP-PIVKA-II inversion, platelet-altitude, and hemoglobin-altitude dimensions. Bayesian evidence integration employed the Effective Number of Independent Pieces of Evidence (ENIPE) method ({delta} = 0.504). Results: In QHRCH-HCC, altitude correlated positively with PIVKA-II ({rho} = +0.244, p = 0.0003) and with an altitude-adaptive genetic background score ({rho} = +0.517, p = 5.59x10-49). Under hypoxia, EPAS1 expression in high-altitude-adapted iPSC-ECs was reduced to 61.4% of controls (p = 0.0006), while STC2 remained relatively unaffected (89.2%, p = 0.180). In TCGA-LIHC, EPAS1[-&gt;]STC2 was weak ({rho} = 0.092) compared with HIF1A[-&gt;]STC2 ({rho} = 0.379, p = 2.21x10-14), establishing a negative control. Cross-cancer validation revealed strong EPAS1[-&gt;]STC2 in ccRCC ({rho} = 0.320, p = 3.47x10-14) but not in LUAD or BRCA. In GDSC2, EPAS1 correlated positively with IC50 of all 11 antiangiogenic TKIs (sign test p = 0.0005). Bayesian updating yielded posterior probability 0.970 (Log10BF = 1.99). Conclusions: EPAS1 LoF represents a germline determinant of TKI response, independent of tumor-acquired mutations. The AESI_score and HIF-2 inhibitor belzutifan constitute a predictive biomarker-therapeutic pair for genotype-stratified clinical validation. This hypothesis-generating study establishes a germline determinant framework for TKI resistance; definitive mechanistic validation will require prospective EPAS1 genotype-stratified cohorts (2023-ZJ-786).

Post-acute infection syndromes (PAIS) follow viral syndromes including post-acute sequelae of COVID19 (PASC) which complicates 10-25% of SARS-CoV-2 infections. These syndromes lack precise explanatory mechanisms. We studied 173 human saliva proteomes during respiratory viral syndromes, seeking associations between 44 clinically-relevant protein expression patterns and subsequent sequelae counts. Exploratory models adjusted by extensive clinical annotations found interactions between 23 acutely-responsive proteins and SARS-CoV-2 infection that inversely predicted subsequent neurocognitive sequelae. An overlapping 19 acutely-responsive proteins during any acute respiratory viral syndrome inversely predicted general fatigue-related sequelae. Altogether, 29 proteins, derived from interferon stimulated genes (ISG), were uniformly beneficial, including 13 predictive of both neurocognitive and general sequelae. The proteins suggested both shared early pathobiology and virus-specific protective responses that shaped resolution of acute disease and different PAIS. Acutely elevated protective ISG proteins associated with reduced post-viral symptoms identify investigational starting points for novel mechanisms, diagnostics and therapeutics for PASC and PAIS.

Interstitial lung diseases (ILDs) are heterogeneous pulmonary disorders characterized by chronic inflammation and/or fibrosis. 30-40% of ILD patients develop fibrotic disease that is associated with progressive respiratory decline and poor prognosis, particularly in idiopathic pulmonary fibrosis. Current antifibrotic therapies slow disease progression but do not reverse fibrosis, highlighting the need for improved therapeutic strategies. Robust histopathological evaluation in preclinical models is essential for drug development; however, conventional scoring systems are semi-quantitative, labor-intensive, subject to inter-observer variability, and rely on limited field sampling. Here, we introduce FibroSight, a standalone platform for compartment-resolved quantification of lung remodeling in Sirius Red-stained sections. By integrating deep learning- based structural segmentation with color-based feature extraction, FibroSight enables highly automated whole-lobe analysis without requiring complex computational setup. The platform quantifies complementary remodeling parameters, including parenchymal collagen fraction, parenchymal tissue density, nuclear area fraction, parenchymal airspace fraction, and airway- and vascular-associated remodeling. Validated in the bleomycin-induced fibrosis model, FibroSight-derived metrics strongly correlated with expert Ashcroft scoring and showed stronger associations with histological severity than corresponding outputs from a semi-automated ImageJ-based workflow. The platform further distinguished inflammatory from fibrotic remodeling in influenza-induced lung injury and demonstrated translational proof-of-concept applicability in human ILD biopsy specimens. By enabling scalable, reproducible, and multi-compartment histological quantification, FibroSight provides a practical framework for objective assessment of lung remodeling. This approach expands conventional fibrosis evaluation by integrating fibrotic, inflammatory, airway, and vascular-associated readouts, supporting more precise analysis of disease mechanisms and therapeutic responses in preclinical and translational ILD research.

Background Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is a poorly understood, debilitating multisystem condition. Converging evidence implicates impaired cellular bioenergetics, neuroinflammation and defective neurovascular coupling that may manifest as "virtual hypoxia" only under physiological stress. Methods We performed a single-session multimodal 3T MRI study combining brain volumetry, arterial spin labelling (ASL) and multivoxel proton magnetic resonance spectroscopy under normoxia and two controlled hypoxic challenges (oxygen saturation 87 {+/-} 3%) in 26 ME/CFS patients and 27 age- and sex-matched healthy controls. Results After intracranial-volume normalization, patients showed a reduced brainstem volume (1.46 0.14 vs. 1.55 {+/-} 0.18 % of eTIV; p = 0.013, FDR-p = 0.039), whereas deep grey matter and whole-brain parenchymal fraction did not differ between groups. Whole-brain cerebral blood flow (CBF) rose under hypoxia in both groups (controls +4.8 {+/-} 13.0%, patients +3.7 {+/-} 11.7%), with greater initial inter-individual variability in patients (patient-to-control variance ratio up to 6.94; FDR-p = 0.001). Thalamic lactate-to-creatine (Lac/tCr) ratios increased with hypoxia in controls (FDR-p = 0.028) but were already elevated at normoxia in patients (0.171 vs. 0.135; FDR-p = 0.021) and did not rise further (FDR-p = 0.38). In exploratory analyses, patients showed exaggerated inverse coupling between thalamic total N-acetylaspartate (tNAA/tCr) and white-matter CBF. Conclusions These findings provide in vivo evidence of impaired neuro-metabolic and vascular adaptive capacity in ME/CFS, supporting the virtual hypoxia hypothesis and highlighting candidate imaging markers for stratification that warrant validation.

Background Data-driven type 2 diabetes (T2D) subtypes differ in their risks of dementia and stroke. We examined whether their metabolomic profiles also differed and whether subtype-related metabolic patterns were associated with dementia, stroke, and all-cause mortality. Methods We analyzed NMR-based metabolomic profiles across previously defined T2D subtypes in the UK Biobank. Subtype-related metabolites were summarized using principal component analysis (PCA), and their associations with incident dementia, stroke, and all-cause mortality were examined using Cox models. Attenuation analyses and two-sample Mendelian randomization further assessed subtype-outcome relationships and the potential causal relevance of outcome-associated metabolites. Results Among 7,671 individuals (mean age 59.85 years; 37% female), the first five PCs explained 76.7% of variance in subtype-related metabolites and mainly reflected lipid and lipoprotein signatures. After adjustment for T2D subtype and confounders, the HDL-remodeling PC increased risks of all-cause dementia (HR 1.17, 95% CI 1.08-1.27), VaD (HR 1.18, 95% CI 1.05-1.32), and all-cause mortality (HR 1.16, 95% CI 1.13-1.19). Lower scores on the LDL cholesterol-enriched axis increase risks of all-cause dementia (HR 0.75, 95% CI 0.62-0.91) and mortality (HR 0.76, 95% CI 0.69-0.83). The VLDL/LDL-enriched PC was inversely associated with mortality (HR 0.93, 95% CI 0.88-0.98). No significant stroke results were observed. Adjustment for the PCA-derived metabolomic patterns generally attenuated subtype-outcome associations, MR analyses identified 197 metabolite-outcome associations that remained significant after FDR correction. Conclusions Metabolomic profiling showed that the metabolic signatures differed across data-driven T2D subtypes and highlighted lipid and lipoprotein remodeling as a major metabolic feature associated with dementia, stroke, and all-cause mortality.

Background: The rising burden of metabolic dysfunction-associated steatotic liver disease (MASLD)-associated hepatocellular carcinoma (HCC) underscores the need for innovative therapeutic strategies. Methods: We integrated RNA-seq fusion detection, immunopeptidomics, and proteogenomics to systematically prioritize tumor-specific neoantigen candidates arising from gene fusions in MASLD-HCC. Results: We elucidated a landscape of private, clonally expressed fusions, and identified a previously unrecognized class of predicted phosphorylated fusion-neoepitopes. Cross-tumor proteomic analysis revealed that these phospho-motifs are present across malignancies, providing a broader context for their biological relevance. Importantly, fusion-positive tumors display immunosuppressive microenvironments, highlighting the need for future therapeutic strategies that combine fusion-targeted immunotherapy with approaches that overcome T-cell dysfunction. Conclusions: This study establishes a discovery pipeline and publicly available resource for fusion-derived phospho-neoepitopes in MASLD-HCC. The identified candidates provide a prioritized framework to guide and accelerate rigorous functional immunogenicity testing for future clinical validation.

Acute Respiratory Distress Syndrome (ARDS) remains highly morbid and lacks approved disease-modifying pharmacotherapies. Direct (pulmonary) and indirect (extrapulmonary) insults may initiate biologically distinct early injury programs, but human tissue-level evidence from the first hours is scarce. Here we establish a paired, acellular ex vivo lung perfusion (EVLP) platform using human donor lungs unsuitable for transplantation to model direct (endobronchial) and indirect (perfusate) lipopolysaccharide (LPS) injury within the same donor. We profiled lung tissue proteomes at 4 h post-insult and performed therapeutic nomination by querying proteomics-derived injury signatures against the CLUE L1000 perturbational compendium with independent cross-platform validation. Both models developed histological injury and robust cytokine release. Direct injury preferentially enriched neutrophil degranulation, extracellular matrix remodelling and metabolic reprogramming modules, whereas indirect injury showed prominent complement/coagulation perturbation with greater endothelial activation markers in perfusate. Cross-platform prioritisation converged on tractable signalling and epigenetic axes, including JAK/STAT, PI3K/AKT/mTOR, SYK, CDK and HDAC inhibitor classes - yielding a tiered shortlist for EVLP intervention testing. This intact human lung perturbation platform enables injury-stratified mechanistic inference and therapeutic prioritisation in early lung injury relevant to ARDS.

Background: Positron emission tomography (PET) myocardial perfusion imaging (MPI) provides complementary information on perfusion, myocardial blood flow and ventricular function. While these markers are often considered collectively during interpretation, their quantitative integration with imaging and clinical data into a unified predictive framework remains limited. We developed a multimodal artificial intelligence framework that combines PET polar maps with quantitative imaging and clinical features to improve obstructive coronary artery disease (CAD) detection. Methods: We retrospectively analyzed the multicenter REFINE PET registry. Among 38,682 PET MPI studies from 14 sites, 2,833 patients without known prior CAD underwent invasive coronary angiography within 180 days. Obstructive CAD was defined as >=50% left main stenosis or >=70% stenosis in other major epicardial coronary arteries. We developed a two-stage contrastive learning framework to learn multimodal PET representations from studies without angiographic labels and transfer them to supervised CAD prediction. In Stage 1, PET image and tabular encoders were pretrained on 12,225 PET MPI studies from eight development sites using 15-channel PET polar maps, quantitative PET perfusion, flow and gated functional measures, and clinical variables. In Stage 2, the pretrained encoders and a lightweight classification head were fine-tuned in 968 angiography-labeled patients, using lower encoder learning rates to limit overfitting. The model was externally validated for angiographically defined obstructive CAD detection in 1,865 patients from six independent sites and compared with standard PET MPI metrics. Results: The prevalence of obstructive CAD was 60% in the training cohort (66% male, median age of 70 years [63, 77]), and 55% in the external validation cohort (64% male, median age of 67 years [60-74]). In external validation, the AI model achieved an AUC of 0.85 (95% confidence interval (CI), 0.83-0.87) for obstructive CAD detection and outperformed conventional quantitative PET metrics (all P < 0.001). At a specificity matched to visual summed stress score, the AI model achieved higher sensitivity (89% [95% CI, 87-91] versus 85% [95% CI, 82-87]) and negative predictive value (81% [95% CI, 77-84] versus 73% [95% CI, 69-77]; both p<0.001). The overall net reclassification improvement was 8.9% (95% CI, 4.2-13.6%; p = 0.001). Conclusions: Multimodal contrastive pretraining improved obstructive CAD detection from PET imaging beyond conventional perfusion-based scoring in independent multisite external validation.

Myasthenia gravis (MG) is an autoimmune disease caused by pathogenic autoantibodies against proteins at the neuromuscular junction (NMJ). The diagnosis and clinical management of MG patients largely relies on the detection of antigen-specific autoantibodies targeting acetylcholine receptor (AChR) or muscle-specific kinase (MuSK). Yet a subset of patients remains seronegative for known MG autoantibodies, highlighting a critical need for alternative approaches to identify pathogenic NMJ antibodies. We established a new human in vitro model of the NMJ based on primary human muscle cells that recapitulates key features of the NMJ: differentiation to myotubes, expression of key NMJ proteins and formation of postsynaptic AChR clusters in response to agrin stimulation. The model allows new insights into myogenesis and genetic muscle diseases, and the new muscle cell-based assay (CBA) detected autoantibodies in sera from patients with AChR- and MuSK-positive MG with 96.43% sensitivity and 100% specificity, while healthy control sera showed no reactivity. Incubation with patient sera significantly reduced AChR clustering compared to controls, demonstrating functional pathogenic effects. Thus, we established a physiologically relevant human NMJ model that enables detection and functional characterization of neuromuscular autoantibodies. This novel approach addresses a key limitation of current antigen-specific diagnostics and provides a method for improved detection and characterization of MG antibodies, independent of antigen specificity. One Sentence SummaryWe established a postsynaptic human in vitro neuromuscular junction model to assess binding and pathogenicity of MG autoantibodies. Key messagesO_ST_ABSWhat is already known on this topic?C_ST_ABSCurrent diagnosis of myasthenia gravis (MG) relies largely on the detection of antigen-specific autoantibodies against AChR and MuSK, leaving a clinically relevant subset of patients seronegative. What are the new findings?We established a physiologically relevant human in vitro neuromuscular junction model based on primary human muscle cells and developed a novel muscle cell-based assay (CBA) for the detection of neuromuscular autoantibodies. How might this impact on clinical practice or future developments?The CBA detected autoantibodies in patients with AChR- or MuSK-positive MG with high sensitivity and specificity and demonstrated their functional pathogenic effects on AChR clustering. This antigen-independent approach may improve the detection and functional characterization of MG autoantibodies, particularly in patients who are seronegative in current diagnostic assays. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=130 SRC="FIGDIR/small/743478v1_ufig1.gif" ALT="Figure 1000"> View larger version (38K): org.highwire.dtl.DTLVardef@18ed154org.highwire.dtl.DTLVardef@151036corg.highwire.dtl.DTLVardef@1b7ab34org.highwire.dtl.DTLVardef@1490fe9_HPS_FORMAT_FIGEXP M_FIG C_FIG

Background & AimsHaematoxylin and eosin (H&E) staining remains the diagnostic gold standard for solid cancers, including hepatocellular carcinoma, and is increasingly complemented by genomic profiling for precision medicine. Inferring genomic alterations directly from H&E images could streamline testing, but heterogeneity and biases in human training data limit interpretation of genotype-phenotype associations. Here, we aimed to relate histologic to genomic pathology to provide biological explainability for mutation prediction models and assess the impact of germline variation on model performance. MethodsWe analysed 597 murine liver tumours with matched whole-genome sequencing and histopathology (163,835 image tiles; 22.9 million nuclei). Our controlled in vivo design accounted for germline variation, biological sex, and causal mutagen (N-diethylnitrosamine), removing confounding factors present in human cohorts. We trained and evaluated deep learning and supervised machine learning models to predict germline variation and cancer driver alterations from H&E. ResultsModelling accurately predicted germline and somatic alterations from histology, at both locus-specific and genome-wide scales. Quantitative image analysis revealed an unexpected association between Egfr driver mutations and hepatic steatosis, linking genotype to an interpretable morphological phenotype. While model performance declined when applied to tumours from unrepresented genetic backgrounds, this limitation was biologically informative, revealing strain-dependent differences in tumour evolution, notably the prevalence of whole-genome duplication. ConclusionsMachine learning integration of histological and genomic pathology enables accurate, interpretable inference of genetic alterations from H&E, potentially reducing reliance on costly ancillary molecular assays. Our predictions are supported by human-interpretable biological features, addressing concerns around "black-box" technologies. However, caution is required when applying such methods to samples with a genetic background that, even if closely related, is beyond the genetic horizon of training data.

Echovirus 18 (E18) is a predominant pathogen causing aseptic meningitis in children, and post-E18 infection frequently triggers myasthenia gravis-like autoimmune neurological damage. This pathological process relies on neonatal Fc receptor (FcRn)-mediated IgG transcytosis across mucosal barriers, and FcRn also acts as an essential functional receptor required for E18 attachment and uncoating during host cell invasion. At present, no E18-specific prophylactic vaccine has been clinically approved, and anti-FcRn monoclonal antibodies are the available therapeutics to alleviate autoantibody-mediated tissue injury. We constructed an integrated automated phylogenetic pipeline named evolution_conservation, which enables rapid tracing of the evolutionary position and genetic relatedness of clinical isolates to identify closely related strains from previous outbreaks. Serving as an in silico alternative to animal experiments, this pipeline supports reference-guided vaccine design and longitudinal comparative assessment of vaccine safety and efficacy, facilitates identification of patient populations presenting rare post-viral sequelae, and accelerates clinical trial progression. In this study, we inserted the pre-screened linear epitope E18-3 into a truncated hepatitis B core (HBc) scaffold to generate chimeric virus-like particles (VLPs). A non-invasive floss-based gingival mucosal immunization mouse model was established, with subcutaneous Freunds adjuvant immunization set as the control group. ELISA results confirmed that gingival mucosal delivery of particulate HBc-E18-3 VLPs alone could induce sustained high levels of antigen-specific intestinal IgG in vivo. Drawing on research paradigms of therapeutic neoantigen vaccines for tumor recurrence prevention, the evolution_conservation bioinformatic pipeline and mucosal VLP platform described herein establish an innovative framework for developing antigen-competitive prophylactic and therapeutic vaccines targeting FcRn for myasthenia gravis and autoimmune encephalitis.

MotivationPharmacovigilance relies on accurate extraction of structured biomedical entities and their semantic relationships from scientific literature. However, most biomedical information extraction systems address named entity recognition (NER) and relation extraction as separate tasks trained on corpus-specific architectures, limiting scalability and cross-task knowledge sharing. Recent developments in instruction-tuned Large Language Models (LLMs) offer a promising alternative through unified generative extraction, but robust schema-grounded multitask adaptation for biomedical extraction is still understudied. MethodsThis study proposes a unified multitask instruction-tuned LLM framework that jointly performs biomedical NER and relation extraction across three benchmark corpora to identify chemical, disease, drug entities, as well as chemical-disease relations, drug-adverse event relations, and drug-drug interactions. Two general LLMs, Llama-3.2-3B-Instruct and Qwen3-8B, were fine-tuned using Low-Rank Adaptation (LoRA) under a shared generation interface that extracts both entity pairs and their underlying relation. Zero-shot and fine-tuned configurations were evaluated across all the tasks on their respective held-out test sets. ResultsParameter-efficient fine-tuning substantially improved both entity and relation extraction performance across all tasks and model families. Fine-tuned Qwen3-8B achieved the strongest overall performance with 89.42% micro-averaged entity F1 and 62.32% micro-averaged relation F1. Fine-tuned Llama-3.2-3B achieved 87.63% entity F1 and 58.42% relation F1 despite its substantially smaller parameter count, outperforming the zero-shot 8B model on both tasks. Fine-tuning also reduced structured JSON parse failures from 23.5% to 0.11%, demonstrating stable schema internalization during supervised adaptation. ConclusionSchema-grounded multitask instruction tuning with LoRA provides a robust and computationally feasible framework for unified biomedical information extraction across heterogeneous benchmark corpora. The findings further demonstrate that schema-grounded adaptation is substantially more important than model scale alone for reliable extraction of structured biomedical relations. The gap between NER and relation extraction performance motivates future research on explicit negative-relation supervision and ontology-guided relation extraction.

Predicting disease risk from prior diagnoses is fundamental to clinical decision-making, particularly during health emergencies such as the COVID-19 pandemic, when individuals with long-term conditions may be disproportionately vulnerable to adverse outcomes. Despite intense interest in developing models to predict disease risk from prior diagnoses (1-3), most prediction models do not estimate effects of each prior diagnosis on disease risk conditional on other diagnoses, limiting interpretability and clinical utility. We developed the Comorbidity Risk Score (CRS), trained on 13 million individuals (age 40-69) from linked electronic health record (EHR) datasets of the entire population of England, to predict COVID-19 hospitalisation and 87 other disease outcomes. CRS was trained at close to saturated sample size and precisely estimated the effects of 212 prior diagnoses on the 88 disease outcomes, conditional on all other prior diagnoses. Correlations of CRS effect sizes across outcomes (e.g. 0.76 for myocardial infarction vs. hyperlipidaemia) matched the corresponding genetic correlations (e.g. 0.79 for myocardial infarction vs. hyperlipidaemia), confirming that comorbidity architectures capture disease aetiology. On average, CRS identified 5% of the population with 3.4-fold higher disease risk, including myocardial infarction (4.4-fold), lung cancer (6.5-fold), and COVID-19 hospitalisation (6.3-fold). Using prior diagnoses alone, CRS outperformed state-of-the-art clinical COVID-19 models (4). Furthermore, CRS (N=13 million) substantially outperformed state-of-the-art AI (1) (N=0.5 million) and linear (3) (N=0.5 million) models in predicting disease risk, suggesting that training sample size outweighs model complexity. CRS attained near-perfect transferability across self-reported ethnicities (e.g., Black vs. White: AUROC ratio = 97.3%). Finally, CRS distinguished independently predictive comorbidities from indirect associations, e.g., lipid metabolism disorder was a strong predictor of myocardial infarction risk but not ischaemic stroke, after conditioning on other prior diagnoses. In conclusion, CRS provides a comprehensive resource for understanding the impact of comorbidities on COVID-19 and other future diseases, revealing disease aetiology while enabling powerful prediction of disease risk.

Hypertrophic and dilated cardiomyopathy (HCM and DCM) carry substantial morbidity and mortality, yet diagnosis may be delayed, particularly when presentation is nonspecific. Existing machine-learning approaches to cardiomyopathy phenotyping, genotype prediction, and risk stratification commonly rely on disease-specific, hand-engineered features drawn from echocardiography, cardiac MRI, ECG, or curated clinical variables. We evaluated whether a general-purpose clinical foundation model, CLMBR-T-base, pre-trained via next-clinical-event prediction with no cardiomyopathy-specific supervision, could produce linearly separable embeddings for all three case/control cohorts. Using EHR data from the Penn Medicine BioBank, we constructed cohorts for (1) prediction of a first recorded qualifying HCM/DCM diagnosis at 1-, 3-, and 6-month horizons, decomposed into eventual-versus-never-case and imminent-versus-eventual comparisons; (2) genetic carrier status prediction among diagnosed patients with completed gene panels; and (3) prediction of heart-failure hospitalization, and all-cause mortality as both binary and time-to-event outcomes. Linear probes fitted to frozen embeddings achieved AUROCs of 0.75-0.82 for onset prediction, 0.74-0.75 for genotype status, and Harrell's concordance of 0.65-0.80 for time-to-event outcomes. Decomposing the onset prediction task reveals that the model often misclassifies patients who were diagnosed later as positive, suggesting the patient journey embeddings encode disease state more reliably than care timing. These results suggest that a single, generically pretrained EHR embedding can support multiple clinically motivated prediction problems in CM without disease-specific feature engineering.

Abstract Background: Every non-invasive continuous glucose monitoring (NI-CGM) technology introduced into the landscape faces the same skeptical question, from regulators, clinicians, and competing developers alike: is the candidate signal actually specific to glucose, or does an apparently reasonable accuracy figure simply reflect a model fitting to motion, temperature, calibration offset, or trial-duration artifact? Existing evaluation practice does not answer this question directly. NI-CGM performance is instead reported almost exclusively with metrics inherited from minimally invasive, subcutaneous CGM, the Mean Absolute Relative Difference (MARD), Clarke/Parkes error grids, and ISO 15197-style agreement rates, which were designed for sensors whose glucose specificity is already chemically established and which therefore take specificity as a premise rather than treating it as a result to be demonstrated. Methods: We present a methodology for demonstrating NI-CGM technology glucose specificity during the algorithm-development phase, and illustrate it with a case study based on a quantum-cascade-laser (QCL) photoacoustic NI-CGM device (Neogly) evaluated in the SKAMo-2 free-living clinical trial (eight participants with type 1 diabetes). The methodology combines a white-noise control, a constant-glycemia control, a sensor-ablation control that removes the candidate physical signal while retaining auxiliary covariates, and explicit reporting of the train/test generalization level, so that a reported MARD can be read as evidence of specificity rather than taken on faith. Results: Removing the mid-infrared photoacoustic (PA) signal from the model while retaining all auxiliary sensors (accelerometer, skin temperature, hygrometry, PPG) degraded performance at every generalization level tested, inter-patient MARD rose from 35.0% with the PA signal to 43.1% without it, and intra-experimentation MARD rose from 22.5% to 23.9%, providing direct, internal evidence that the PA channel itself, and not merely the auxiliary covariates, carries glucose-specific information. At the same time, an algorithm trained on pure Gaussian noise produced a MARD of 25% over short test windows, and a trivial constant-glycemia predictor outperformed every machine-learning model tested when generalization was extended from a single recording to an unseen patient (MARD 55% for the naive constant model versus 37% for a deep neural network on inter-patient splits). Reported in isolation, any of these MARD values is uninterpretable; reported against one another, they jointly demonstrate that the signal is specific to glucose while also bounding how much of the headline accuracy figure that specificity currently explains. Conclusions: We propose a specificity-demonstration methodology for NI-CGM technology development, comprising (1) signal quality gating prior to any algorithm benchmarking, (2) a white-noise control to test for genuine information content, (3) a constant-glycemia control to expose trial-duration bias, (4) a sensor-ablation control that isolates the contribution of the candidate physical signal from auxiliary covariates, (5) explicit reporting of the data-splitting generalization level (intra-experimentation, intra-patient, inter-patient). This methodology answers a question that precedes clinical accuracy reporting and that recognized clinical frameworks such as the IFCC Working Group on CGM's Dynamic Glucose Regions guideline are not designed to answer: not how accurate is the device, but is the device measuring glucose at all. We argue that without these controls, MARD and error-grid values for NI-CGM are not comparable across studies and may either overstate clinical readiness or undermine promising technologies. We recommend that this specificity methodology be applied routinely once a candidate NI-CGM sensor reaches algorithm-development stage, alongside and as a deliberate complement to IFCC-style clinical accuracy reporting once the device is mature enough for that evaluation. Keywords: non-invasive continuous glucose monitoring; glucose specificity; algorithm validation; MARD; benchmarking; machine learning; photoacoustic spectroscopy; sensor ablation; Clarke error grid