Advanced Science

Severe motor impairments such as amyotrophic lateral sclerosis and locked-in syndrome lead to partial or complete loss of speech, severely restricting communication as voluntary motor control deteriorates. In this study, we developed a non-invasive, wearable EEG-based brain-computer interface that reconstructs coherent natural language sentences by decoding linguistic components directly from neural activity. EEG data were collected from 20 healthy volunteers using two bilateral temporal electrodes (T7/T8) sampled at 128 Hz, with signals segmented into 4-second windows (512 samples) and normalized to a [-1, 1] range. Participants silently generated 6 pronouns, 80 verbs, and 217 object nouns across six languages under controlled cognitive tasks. Separate 1D convolutional neural networks were trained to classify pronouns and to regress verbs and nouns into 300-dimensional FastText semantic embeddings using cosine similarity loss, with data splits of 70/30 or 80/20 for training and validation and augmentation applied exclusively to training data. Pronoun decoding achieved over 60% training accuracy but showed reduced generalization (validation AUC {approx} 0.55), reflecting the context-dependent nature of indexical language, while verb and noun models demonstrated stable convergence over 200-300 epochs and successfully mapped EEG features into semantic space. A Siamese network integrated decoded components into a shared embedding space to ensure semantic coherence prior to sentence generation, enabling the production of grammatically correct and contextually appropriate sentences aligned with experimental conditions such as hunger and thirst. Validtaion cohort of 10 individuals was used to predict their thoughts related to hunger, while fasting. 80% accuracy was achieved. These findings demonstrate that bilateral temporal EEG signals alone are sufficient to recover structured linguistic intent when combined with similarity-based semantic modeling, advancing a scalable, non-invasive communication framework for clinical translation.

BackgroundMenopausal obesity is a type of obesity in women during menopause where the decline of ovarian function and the decrease of estrogen levels lead to an imbalance between energy intake and consumption in the body, resulting in fat accumulation and weight gain. Moxibustion, as a green therapy of non-interventional external treatment that prevents and treats diseases through thermal stimulation of relevant acupoints, has been widely used in clinical practice because of its simplicity, convenience, effectiveness, low price and high compliance. PurposeTo clarify the pathogenesis of menopausal obesity and the biological mechanism of moxibustion treatment for menopausal obesity. MethodsWe selected 9 plasma samples from menopausal obese patients before and after moxibustion treatment, as well as 9 plasma samples from the healthy control group. After sample mixing and replication, DIA quantitative proteomics analysis was used to screen out differentially expressed proteins, and bioinformatics analysis was conducted. ResultsThe plasma proteomic analysis revealed a significant increase in the protein expression levels of APOC2 and PZP in menopausal obesity patients. These differential proteins primarily participate in biological regulation, cell metabolism, and reproductive development processes. Their biological processes and molecular functions are mainly associated with enzyme inhibitor activity, calcium-dependent protein binding, lipid localization, and plasma lipoprotein particle assembly. The pathogenesis of menopause obesity is linked to the accumulation of visceral fat resulting from changes in sex hormone levels and reduced energy consumption following the decline of female ovarian function. Following moxibustion treatment, there was a notable down-regulation in the plasma levels of sialoglycoprotein receptor 2 (ASGR2), membranin A1 (ANXA1), and human heterogeneous nuclear ribonucleoprotein C (HNRNPC) among menopausal obesity patients. Their biological processes and molecular functions were primarily concentrated on intracellular hagy, nucleic acid binding, tissue regeneration, and neutrophil clearance. ConclusionThe mechanism underlying moxibustions effectiveness in treating menopausal obesity may involve down-regulating HNRNPC expression, activating the PI3K/Akt/mTOR autophagy signaling pathway, regulating hormone levels to delay ovarian aging thereby improving lipid metabolism.

BackgroundWhile blood-based biomarkers for Alzheimers Disease (AD) such as p-Tau and NfL characterize established pathology, the systemic biological cascade triggering these events remains incompletely mapped. We hypothesized that proteins exhibiting a rising trajectory in the prodromal phase might reveal novel mechanisms of disease progression. MethodsUsing data from the UK Biobank Pharma Proteomics Project (N = 4,519 incident AD cases), we performed a blind trajectory scan of [~]3,000 plasma proteins. We utilized an elimination strategy, systematically excluding known AD markers (e.g., APOE, NEFL) and verified biological responses (e.g., MMP3, GLRX) to isolate novel signals. ResultsAfter excluding established markers, VSIG10L--a V-set and immunoglobulin domain-containing protein--emerged as the most significant novel marker (beta = - 0.037, P = 0.0019), exhibiting a progressive rise as patients approached diagnosis. Crucially, VSIG10L was accompanied by a cluster of co-regulated proteins involved in embryonic development and cell cycle regulation, including NACC1 (stem cell pluripotency), VASN (vasculogenesis), and ZBTB17 (cell cycle checkpoint). ConclusionThe emergence of VSIG10L and its associated developmental cohort suggests that prodromal AD is characterized by a retrogenesis phenomenon, the unsilencing of developmental programs in a failed attempt at neural repair. These proteins offer a new window into the brains response to neurodegeneration and represent potential therapeutic targets.

BackgroundRetrieval-augmented generation (RAG) frameworks such as RAPID [1] have demonstrated that staged planning and retrieval grounding improve long-form text generation. However, most implementations remain similarity-driven and open-domain, lacking the epistemic safeguards required for biomedical synthesis, where mechanistic completeness, temporal governance, traceability, and explicit gap classification are essential. ObjectiveTo develop and evaluate a topology-aware, graph-augmented retrieval framework for structured biomedical narrative synthesis, and to position it as a domain-constrained evolution of staged RAG aligned with structural principles of digital evidence-based medicine (dEBM). MethodsWe implemented a two-layer architecture operating on a closed, version-controlled corpus of 11,861 peer-reviewed text chunks on iron deficiency. A metadata-constrained vector retriever (RAG01) was extended with a Graph-RAG (RAG02) overlay (RAG02) constructed from chunk-level entity extraction and weighted co-occurrence networks (30 nodes; 118 directed edges). Topic planning was organized through predefined mechanistic axes functioning as structured hypothesis probes. Retrieval was performed under identical deterministic constraints (top-k = 5; cosine threshold = 0.50; publication year [≥] 2023), and graph diagnostics--including local connectivity, induced subgraph density, modular overlap, and multi-hop stability--were used to distinguish retrieval insufficiency from corpus-level evidentiary scarcity. ResultsIn a case study of obesity-associated iron deficiency, the entity network exhibited a centralized regulatory topology with hepcidin as a high-connectivity hub. Axis-based retrieval combined with graph auditing consistently reinforced an inflammation-mediated hepcidin pathway linking obesity to iron deficiency, while alternative mechanisms lacked stable multi-hop embedding. Compared with vector-only retrieval, graph augmentation preserved semantic alignment and increased mean cosine similarity from 0.673 to 0.694 while reducing similarity dispersion (SD 0.056 to 0.035) under identical constraints. Graph activity ratio was 1.00 in the temporally filtered corpus. ConclusionsBy integrating mechanistic axis decomposition, topology-aware auditing, causal scaffolding, and expert-driven iterative refinement, the proposed framework implements selected structural constraints inspired by evidence-based medicine within a controlled digital synthesis environment. The approach advances retrieval-augmented generation beyond similarity-based summarization toward a reproducible model of topology-aware biomedical evidence interrogation with implications for AI-assisted systematic reviews.

Biological age estimators quantify aging-related variation but provide limited insight into organ-specific aging processes. The retina enables non-invasive visualization of microvascular and neural structures and has emerged as a promising modality for biological age prediction. However, existing retinal aging models typically produce unidimensional age estimates with limited interpretability. Here we develop a deep learning framework based on a large-scale vision foundation model to estimate retinal biological age from fundus images and to characterize the physiological heterogeneity underlying retinal aging. Using a reference cohort of 56,019 relatively healthy individuals, the model achieved a Mean Absolute Error of 2.48 years in age prediction. Analysis of age deviations in a real-world clinical cohort (n = 46,369) revealed non-linear associations with cardiometabolic risk and population heterogeneity in aging patterns. Integrating multidimensional physiological profiling, feature attribution and unsupervised analysis, we identified distinct retinal aging signatures associated with systemic inflammation and hemodynamic variation. To further characterize age-related deviations, we introduced a residual learning framework that decomposes retinal aging signals into a normative age-related component and additional components associated with physiological variation, achieving a Mean Absolute Error of 1.80 years on the independent healthy test set. This approach provides an interpretable representation of retinal aging and a framework for studying organ-level aging processes and their relationship to systemic health using large-scale imaging data.

Temporal interference stimulation (TIS) promises deeper and more selective neuromodulation, yet predictions remain sensitive to uncertainties in electrode setup and head modeling. We investigate the impact of coregistration error (CE) of the volume conductor and the head, electrode placement uncertainty (EP), and tissue conductivity uncertainty (CU) on the electric field generated by TIS. The stochastic model aggregates CE, EP, and CU into nineteen random variables and is evaluated for a deep target in the left hippocampus and a superficial target in the motor cortex. The uncertainty and sensitivity analysis of the maximal modulation envelope of the electric field is based on an adaptive polynomial chaos expansion (PCE). Spatial statistics show that the mean of the electric field remains focalized over the regions of interest (ROIs), whereas the standard deviation is concentrated in targeted regions, indicating that uncertainty perturbs the electric field magnitude more than its focality. Variance decomposition reveals a clear hierarchy: CU is the main contributor to field variability, EP has a modest influence, and CE is essentially negligible within the considered ranges. Probability-density estimates of the mean strength inside and outside ROIs demonstrate separated distributions, confirming strong dose selectivity for both deep and superficial targets. Overall, within realistic modeling and setup uncertainties, TIS targeting appears robust when using state-of-the-art MRI-based electrode localization. The analysis identifies insufficient knowledge of tissue conductivities as the primary limitation for further improving the reliability of electric field predictions in TIS.

Background & AimsPerihilar cholangiocarcinoma is an aggressive malignancy with clinical heterogeneity and poor long-term outcomes after resection. Current prognostic assessment relies mainly on anatomical staging and pathological features, which incompletely capture the entire postoperative risk. We aimed to determine whether integrative analysis of clinical, surgical, pathological and tumor genomic data could improve time-resolved, individualized recurrence-risk prediction after curative-intent resection. MethodsWe performed a multicenter retrospective study including patients undergoing curative-intent resection for perihilar cholangiocarcinoma in ten Spanish hospitals (2003-2023). Overall and disease-free survival were analyzed using Cox models. Outcome-agnostic clinical phenotypes were derived by unsupervised clustering of clinical and surgical features. Targeted tumor sequencing of cancer-associated hotspot regions and selected genes was performed. Prognostic models integrating clinical and genomic data were trained and evaluated in independent training/test sets using penalized and latent-component Cox frameworks, with time dependent discrimination. ResultsThe final cohort comprised 142 patients, with a median follow-up of 26.4 months. Recurrence occurred in 61.3% of patients, and 53.5% died during follow-up. Classical pathological factors were strongly associated with survival and recurrence. Unsupervised outcome-agnostic clustering identified three reproducible clinical phenotypes with markedly different recurrence patterns and survival, only partially explained by anatomical staging. Integrative clinical-genomic modelling further improved recurrence-risk prediction, achieving high discrimination in independent validation (time-dependent AUC [~]0.8). Moreover, the integrative model assigned higher risk over time to patients who relapsed. Patients combining unfavorable clinical phenotype with high genomic-derived risk exhibited a high probability of early recurrence. ConclusionsIntegrated clinical phenotyping and targeted genomic profiling substantially refine recurrence-risk stratification after resection of perihilar cholangiocarcinoma beyond anatomical staging alone. This provides a pragmatic framework for risk-adapted postoperative surveillance and therapeutic decision-making. Impact and ImplicationsThis study provides a data-driven framework integrating clinical, surgical and targeted genomic information to refine prognostic stratification after resection of perihilar cholangiocarcinoma, addressing the limitations of anatomy-based staging in capturing biological heterogeneity. The results are particularly relevant for clinicians managing postoperative surveillance and adjuvant strategies, as they identify patient subgroups with markedly different risks of early recurrence despite similar conventional staging. In practical terms, the combination of unsupervised clinical phenotyping and a targeted, biologically informed genomic panel could support risk-adapted follow-up intensity, selection for adjuvant or experimental therapies, and enrolment into clinical trials. While prospective validation is required before routine implementation, this approach offers a feasible and interpretable pathway toward precision postoperative management in a highly aggressive malignancy.

ImportanceCognitive impairment (CI) is often under detected in primary care due to time and resource constraints. Passive analysis of clinical dialogue may offer an accessible approach for screening. ObjectiveTo assess whether audio recordings of patient-physician dialogue during routine primary care visits can be used to identify CI using acoustic speech features and machine learning (ML). DesignThis observational study conducted among older primary care patients involved audio recording primary care visits using a microphone and portable device. An external validation cohort was recruited in a separate city to assess reproducibility of findings. SettingThe study was conducted in primary care practices in New York City, with additional participants recruited from primary care practices in Chicago, Illinois, for validation. ParticipantsThe study included 787 English-speaking patients aged 55 years and older, without documented history of dementia or mild CI. Eligible patients were recruited from primary care practices during routine visits. For validation, 179 patients meeting the same eligibility criteria were recruited from primary care practices in Chicago. ExposuresMultiple thirty-second speech segments were extracted from recordings. Acoustic features were derived using foundation models (Whisper, HuBERT, Wav2Vec 2.0) and expert-defined methods (eGeMAPS, prosody). Main Outcomes and MeasuresCI was defined as Montreal Cognitive Assessment score [≥]1.0 standard deviations below age and education-adjusted norms. ML classifiers were trained to predict CI status from audio recordings. We calculated area under the receiver operating characteristic curve (AUC-ROC) and maximum F1 score (Fmax) for identifying CI participants. ResultsThe mean age was 66.8 years and 21% had CI. Models using Whisper-derived acoustic features performed best (AUC-ROC=0.733, 95% confidence interval [95%CI]=0.714-0.752; Fmax(CI)=0.504, 95%CI=0.474-0.534). Results generalized to the external site with similar performance (AUC-ROC=0.727, 95%CI=0.714-0.740; Fmax(CI)=0.459, 95%CI=0.442-0.476). Model interpretation identified pitch, timing, and variability features as key predictors. When used for screening, the algorithm achieved positive predictive value of 30.4% (95%CI=28.7%-32.1%), sensitivity of 68.2% (95%CI=61.8%-74.6%), and specificity of 63.6% (95%CI=59.8%-67.4%) on the holdout cohort. Conclusions and RelevanceML models trained on acoustic features from brief clinical conversations identified CI with high accuracy. These findings support the feasibility of passive, speech-based screening during routine primary care. Key Points QuestionCan acoustic features extracted from audio recordings of patient-physician conversations during routine primary care visits be used to screen for cognitive impairment? FindingsIn this study including 787 older adults without diagnosis of cognitive problems, machine learning models trained on acoustic features from audio segments of recordings of primary care visits achieved area under the receiver operating characteristic curve values of 0.72 for predicting cognitive impairment. The algorithm achieved a sensitivity of 83%, specificity of 44%, and positive predictive value of 28%, identifying a subset of primary care patients at higher risk for cognitive impairment. Models performed similarly on an external validation dataset of 179 participants. Interpretability analyses highlighted patient pause duration and energy-related features as salient indicators of cognition status. MeaningThese findings suggest that short segments of naturalistic clinical dialogue may contain useful acoustic signals for passively screening patients for cognitive impairment.

Glycated hemoglobin (HbA1c) is a central biomarker for long-term glycemic control and diabetes management, traditionally quantified using laboratory-intensive chromatographic or immunochemical assays. As the global burden of diabetes continues to rise, there is growing interest in alternative, scalable approaches capable of rapid biochemical assessment. Fourier-transform infrared (FTIR) spectroscopy offers a reagent-free method that captures molecular signatures of protein glycation, but translating complex spectra into clinically interpretable HbA1c values requires robust analytical frameworks. Here, we present a complementary multi-model strategy for predicting HbA1c from FTIR spectra of whole blood. Using 685 blood samples with matched reference HbA1c measurements, we evaluated three analytically distinct yet synergistic approaches: partial least squares regression (PLSR), peak-resolved curve fitting based on pseudo-Voigt functions combined with H2O AutoML, and a convolutional neural network (CNN). PLSR and CNN models were trained on biologically informative spectral regions (800-1800 cm-{superscript 1} and 2800-3400 cm-{superscript 1}), while curve fitting focused on the fingerprint region (1000-1720 cm-{superscript 1}) to extract interpretable biochemical parameters. PLSR achieved the highest predictive accuracy (R{superscript 2} = 0.76), closely followed by the CNN (R{superscript 2} = 0.73), reflecting their ability to capture global linear and nonlinear spectral relationships. Although curve fitting yielded lower predictive performance (R{superscript 2} = 0.59), its peak-level decomposition enabled mechanistic interpretation of glycation-related changes. Explainable AI analysis using SHAP identified lipid- and protein-associated vibrations, carbohydrate-linked glycation bands, and amide-region structural features as key contributors to HbA1c prediction. Rather than treating these approaches as competing alternatives, our results demonstrate that their integration provides a more informative framework than any single model alone. By combining predictive performance with biochemical interpretability, this multi-model FTIR strategy highlights a scalable and mechanistically grounded pathway toward non-invasive HbA1c assessment and broader metabolic screening in diabetes monitoring. The code for this study is freely available at https://github.com/MelnychenkoM/ftir-hba1c-prediction.

The daily cortisol cycle is a critical indicator of hypothalamic-pituitary-adrenal (HPA) axis function. The current analytical approaches produce several outputs difficult to integrate into simple statistical models, clinical workflows, and ML/AI pipelines requiring single-value inputs. We developed the Cortisol Sine Score (CSS), a model-free scalar metric that quantifies daily cortisol exposure by computing a weighted sum of cortisol measurements across the day, using sine-transformed time-of-day weights. The CSS produces positive values for morning-dominant patterns, negative values for evening-shifted profiles, and near-zero values for flattened rhythms characteristic of chronic stress and circadian disruption. We validated the CSS performance in 3,006 samples from 501 pregnant women enrolled in the March of Dimes program, with cortisol values measured at 6 time points per day collected during the second trimester of pregnancy. The CSS showed strong correlations with observed and model-estimated amplitude and acrophase from Cosinor regression and JTK_CYCLE approaches, with excellent classifying performance (AUC=0.89, high versus low). The CSS successfully captured established associations between social disadvantage and cortisol dysregulation, and demonstrated utility in predicting gut microbiome composition in metagenomic analyses. Importantly, the CSS maintains excellent fidelity to the full 6-sample protocol with as few as 3-4 daily measurements. The 4-sample protocol achieves great performance (r = 0.952, MAE = 0.087) while reducing participant burden. The 06:00 time point was identified as essential for accurate CSS quantification. The CSS bridges the gap between circadian analysis and practical implementation by providing a simple, interpretable, and robust assessment of cortisol daily cycle in large-scale epidemiological studies, clinical screening, and biomedical sensors. HighlightsO_LICurrent state-of-the-art approaches estimating the daily cortisol exposures produce multi-output information difficult to implement in simple statistical analyses or ML/AI multi-omics approaches C_LIO_LICortisol Sine Score is a novel model-free scalar metric expressing cortisol daily exposure and rhythmicity (morning vs evening exposure) C_LIO_LICortisol Sine Score was validated using 3006 salivary samples from clinical data and golden standards in circadian analyses such as Cosinor and JTK_CYCLE C_LIO_LICortisol Sine Score was the top performer in our benchmarking approach predicting association with social disadvantage and gut microbiome composition C_LIO_LIReliable with 3-4 daily samples, reducing participant burden C_LIO_LIOpen-source R package CortSineScore democratizes cortisol cycle analysis C_LI

Kidney transplantation faces a critical paradox: while thousands await organs, approximately 30% of potential deceased donor kidneys are discarded for various reasons, including subjective assessments due to the lack of an objective molecular biomarker of preservation quality. Here, we applied novel "top-down" proteoform imaging mass spectrometry across living donor (LD), deceased donor (brain death or cardiac death), and discarded human kidneys to quantify proteoforms correlating with post-transplant kidney function. This approach preserves post-translational modifications and splice variants, revealing molecular tissue variability beyond protein presence. LD kidneys displayed robust metabolic signatures, including L-xylulose reductase and cytochrome oxidase subunits, whereas deceased donor and discarded organs showed elevated cellular stress markers such as alpha-B-crystallin and peroxiredoxin 1. Post-transplant blood proteoform analysis validated tissue findings, demonstrating persistent cellular stress and immune activation in deceased donor recipients compared with physiologic wound healing in LD recipients. Consistent with these molecular predictions, serum creatinine levels were highest in DCD, intermediate in DBD, and lowest in LD recipients. The intersection of tissue proteoform signatures across all marginal tissues identified four proteoforms consistently elevated in deceased and discarded kidneys: ACTG1, acetylated CRYAB, PARK7, and S100A4. Collectively, these proteoforms capture key molecular indicators of graft quality, reflecting oxidative stress, cellular injury, and immune activation pathways. As such, they represent promising point-of-care (POC) biomarker candidates for objective kidney classification, potentially improving donor kidney utilization. Translational statementCurrent methods for evaluating donor kidney quality rely on subjective assessments, contributing to the discard of approximately 30% of potentially viable organs. This study demonstrates that "top-down" proteomics can objectively identify molecular signatures distinguishing high-quality from marginal donor kidneys. Top-down proteomics analyzes intact proteins with their post-translational modifications or cleavage products, termed proteoforms to provide mechanistic insights into graft quality. We identified four proteoforms (ACTG1, acetylated CRYAB, PARK7, and S100A4) to be consistently elevated in deceased and discarded kidneys, reflecting oxidative stress, cellular injury, and immune activation. These molecular markers correlated with post-transplant kidney outcome, as measured by serum creatinine levels and recipient blood proteoforms. As a next step, validation in larger cohorts could establish these proteoforms as point-of-care biomarkers for real-time donor kidney assessment during procurement. This objective molecular stratification could reduce unnecessary organ discards and improve transplant outcomes by matching organ quality with recipient risk profiles.

Population aging heightens the burden of cognitive decline and brain disorders, yet trajectories of brain aging vary widely across individuals. Because the human brain is intrinsically lateralized, age-related shifts in hemispheric asymmetry may reveal latent aging subtypes that are masked by bilateral averages. Here, we derived reproducible and interpretable asymmetry-based brain-aging modes and validated their behavioral, genetic, and molecular signatures. Using UK Biobank MRI, we computed cortical-thickness asymmetry across 68 Desikan-Killiany regions, transformed signed asymmetry into non-negative channels, and assembled a region-by-participant matrix. We then applied non-negative matrix factorization (NMF) to estimate spatial mode maps and participant-specific loadings, selecting the factorization rank by reconstruction-error elbow criterion (k = 13). Age associations were assessed with covariate-adjusted partial correlations controlling sex and handedness and corrected for multiple testing using false discovery rate (FDR). Generalizability was evaluated by projecting an independent cohort (Cam-CAN; n = 608) onto UK Biobank-derived spatial maps. We additionally tested sex differences, lifestyle/behavioral correlates, transdiagnostic polygenic risk score (PRS) coupling across 12 neuropsychiatric/neurodegenerative disorders, and imaging-transcriptomic pathway enrichment using Allen Human Brain Atlas expression and Metascape. We identified five age-linked asymmetry modes that replicated directionally in Cam-CAN. Modes differed systematically by sex and displayed distinct lifestyle signatures spanning sleep, physical activity, alcohol intake, diet, device use, and smoking. Genetic coupling was mode-specific, with different modes aligning with distinct constellations of transdiagnostic PRS. Imaging-transcriptomic analyses further indicated mechanistic dissociability, implicating mitochondrial bioenergetics, antigen presentation, innate immune/inflammatory pathways, and synaptic/neurodevelopmental programs. Hemispheric asymmetry decomposes into reproducible, mechanistically diverse aging modes that connect to modifiable behaviors and transdiagnostic genetic liability. This asymmetry-informed, mode-based framework advances subtype-oriented phenotyping of brain aging and provides a foundation for individualized risk stratification and mechanistic hypothesis generation.

BackgroundParkinsons Disease (PD) is often reduced to its most visible motor symptoms, yet it is a systemic neurodegenerative disorder with a highly heterogeneous presentation. While cardinal motor signs such as bradykinesia, rigidity, and tremor arise from the loss of dopaminergic neurons in the substantia nigra, they typically manifest only after substantial neurodegeneration (approximately 50-70% loss) has already occurred, inevitably leading to delayed detection [1] PD significantly impacts non-motor and fine-motor domains as well that are frequently overlooked. Research indicates that hypokinetic dysarthria (voice impairment) affects approximately 89% of PD patients, often as an early prodromal sign [18]. Similarly, micrographia (handwriting impairment) is observed in up to 63% of cases, while non-motor symptoms such as hyposmia (loss of smell) and REM sleep behavior disorder occur in over 70% and 40% of patients, respectively--often years before clinical diagnosis [19, 20]. Consequently, diagnostic systems that rely on a single modality fail to capture this complexity, leading to missed detections in patients whose primary symptoms fall outside that specific domain. To address this, we propose a holistic, multimodal AI framework that explicitly targets these diverse pathological vectors--Voice, Gait, Handwriting, and Non-Motor Symptoms--to ensure robust and early detection across the full spectrum of the disease. MethodsWe propose a modular multimodal AI framework that integrates five complementary inputs: voice recordings, signals captured with the help of a smart pen during drawing spiral/meander, hand-drawn spiral/meander images, wearable sensor-driven gait data, and MDS-UPDRS questionnaire-derived symptom scores. Each modality undergoes an independent preprocessing and specialized modeling pipeline. Outputs from these specialized models are combined using a weighted aggregation engine, which allows for customizable contribution of each modality to the final classification. ResultsPreliminary experiments show that the unimodal pipelines achieved high accuracy, with the Random Forest (Voice) achieving 89%, XGBoost (Drawing Signal) up to 93%, and ResNet-18 (Drawing Image) up to 92%. Incorporating the Transformer model for gait data, which achieved 86% accuracy, significantly boosts the detection of subtle motor deficits. The proposed approach is expected to improve the overall diagnostic sensitivity and specificity relative to any unimodal baseline, offering transparent score breakdowns for clinical use. ConclusionThis study validates a comprehensive, multimodal Machine Learning framework designed to capture the holistic nature of clinical Parkinsons Disease. Our results indicate that fine motor control--analyzed through both dynamic handwriting signals and static imagery--serves as a highly discriminative biomarker, offering superior detection of subtle kinematic tremors. Furthermore, the integration of vocal analysis and spatiotemporal gait modeling ensures that the system captures the full spectrum of pathology, distinguishing between phonatory deficits and gross motor irregularities. By synthesizing these diverse clinical indicators, the proposed architecture overcomes the sensitivity limitations of single-modality systems, establishing a robust, non-invasive foundation for objective early screening and longitudinal patient monitoring in real-world settings.

Gait variability is a critical functional indicator of dynamic balance and neurocognitive decline in health. Its translation into clinical practice is, however, challenged by a lack of age-related normative trajectories and reference values under real-world ecological settings. Furthermore, the conventional metrics used to estimate gait variability (Coefficient of Variation, CV; Standard Deviation, SD) have a fundamental methodological flaw: the inherent sensitivity of conventional metrics to the statistical outliers and environmental noise in real-world walking. In this study, we mitigate this factor by applying a robust statistical framework to quantify gait variability. Analysing a large-scale cohort of community-dwelling older adults (n=2,193), we first demonstrate that free-living gait data follows a heavy-tailed distribution, necessitating the use of robust estimators like the Robust Coefficient of Variation (RCV-MAD) and Median Absolute Deviation (MAD). Leveraging these metrics, we established the normative trajectory and reference values of real-world gait variability across the ageing lifespan, revealing a distinct, age-dependent increase in spatio-temporal fluctuations, indicating a decline in rhythmicity and steadiness with age. We further demonstrated the clinical utility of these robust metrics: RCV-MAD consistently yielded larger effect sizes than conventional CV in discriminating between fallers and non-fallers across all gait parameters. Furthermore, we illustrate the potential of long-term unsupervised monitoring to capture intrinsic variability during real-world walking. Validated for consistency and reliability, this robust framework provides the necessary ecological validity to transform gait variability into a standardised, rapid clinical metric for assessing functional decline at an early timepoint.

BackgroundIn lung transplantation, de novo immunodominant donor-specific anti-HLA antibodies recognizing HLA-DQ antigens (dn-iDSA-DQ) are predominant and can induce chronic lung allograft dysfunction (CLAD). We previously developed a method to measure the active concentration of dn-iDSA-DQ. We aimed to determine whether this new quantitative biomarker is associated with transplantation outcomes. MethodsThis retrospective multicentre cohort study included 90 lung transplant recipients (LTRs) developing dn-iDSA-DQ, evidenced through single antigen flow beads (SAFB) follow-up. We measured the active concentration of dn-iDSA-DQ at the time of their first detection (T0) for all LTRs, and within the 2 years after DSA detection, whenever possible. SAFB dn-iDSA-DQ characteristics and clinical data were retrieved up to 5 years after DSA detection. ResultsWe tested 184 sera with SPR (n=90 at T0, n=94 within the 2 years after DSA detection), among which 63 (34.4%) had a quantifiable concentration of the dn-iDSA-DQ ([&ge;]0.3 nM). The median SAFB mean fluorescence intensity (MFI) of the dn-iDSA-DQ with a concentration [&ge;]0.3 nM was higher (p<0.0001), yet the correlation between SAFB MFI and active concentration was low (r=0.758, p<0.0001). In multivariate analysis, a concentration of the dn-iDSA-DQ [&ge;]0.3 nM at T0 was independently associated with a lower 2-year CLAD-free survival (HR 2.06, p=0.02). A concentration of the dn-iDSA-DQ [&ge;]0.3 nM within the 2 years from DSA detection was associated with a lower graft survival in univariate analysis. ConclusionsActive concentration of dn-iDSA-DQ appears as a valuable biomarker to identify pathogenic DSA at their first detection because of its association with CLAD.

Human muscle biopsies are often required to study or diagnose diseases. However, traditional approaches are challenging due to limited sample size, quality, or patient discomfort. Fine-gauge needle biopsies ([&ge;]14-gauge), present an alternative but yield insufficient sample sizes for histology or function. Ultrasound guidance, coupled with vacuum-assisted, single needle-insertion multiple sampling addresses these challenges. In 19 healthy participants (mean age: 30.1{+/-}10 years, 42% male), 2-3 samples were collected from a single needle insertion into the vastus lateralis (VL) and tibialis anterior (TA). Summed VL and TA sample masses averaged 148{+/-}38mg and 166{+/-}64mg, with dimensions of 15.83{+/-}8 x 2.9{+/-}0.6mm2 (VL) and 15.07{+/-}7 x 3.1{+/-}0.9mm2 (TA). VL had a mean fiber cross-sectional area of 4,347{+/-}1,931{micro}m2, with 221{+/-}86 fibers quantified. Samples were of sufficient size and quality for thorough analyses from a single biopsy procedure, including mitochondrial respirometry, RT-PCR, collagen content, and biomechanical function. Fibers produced typical isometric stress values of 187kPa with a passive modulus of 239kPa (peak) and 79kPa (stress-relaxed). This procedure was well tolerated, with an average immediate pain rating of 1.5{+/-}1 (range:0-4, scale: 1-10) and 24-hour follow-up rating of 1.7{+/-}1 (range:0-4). This report describes an approach that yields high-quality muscle samples suitable for histological and biochemical analyses while minimizing discomfort.

BackgroundPhenotype Risk Scores (PheRS) leverage electronic health record (EHR) data to identify individuals at risk for Mendelian disorders, but their performance remains untested for diseases with common and/or non-specific features such as variant transthyretin amyloidosis (ATTRv), often presenting with heart failure (HF), atrial fibrillation, polyneuropathy, and other prevalent diagnoses. We optimized a PheRS for the most common form of ATTRv by integrating genomic and clinical data in Mount Sinais BioMe biobank, focusing on expert-driven phenotype definitions for the TTR variant p.Val142Ile (V142I), which is prevalent in African American (AA) populations (4%). MethodsWe developed and evaluated a customized PheRS for ATTRv that incorporated 21 expert-curated phenotypic features including 292 ICD-9 and ICD-10 diagnosis codes on a biobank cohort of V142I+ cases (n=383) and controls without any pathogenic/likely pathogenic TTR variants (n=30,642). We compared its performance with the standard automated PheRS approach using different metrics. To account for age-dependent penetrance and high lifelong risk of HF, we further tested the customized PheRS for V142I in a subset of individuals of age [&ge;] 60 with self-reported Black or AA race/ethnicity and at least one occurrence of HF in their EHRs. ResultsThe expert-curated PheRS outperformed the standard PheRS as measured by improved precision-at-k (0.05 vs. 0.00; k=100), a demonstrably, clinically relevant metric. In the subcohort enriched for anticipated penetrance (older, Black/AA HF patients), the expert-curated PheRS identified more V142I+ individuals (6.0%) among the top 100-scoring individuals than a strategy that randomly sampled from the population (3.6%). ConclusionThis work demonstrates that standard PheRS methods are insufficient for common, adult-onset cardiovascular genetic diseases such as V142I-related ATTRv, but when redesigned with disease biology, ancestry, age, and clinical context in mind, PheRS become clinically actionable tools for precision cardiology.

The exposome factors, such as diet, lifestyle, microbiome, chemical exposures and social exposome, shapes human health beyond genetic influences, but the mechanisms remain only partially understood. Leveraging the Area Deprivation Index (ADI) of Neighborhood Atlas, a validated measure of the US social exposome, we derive molecular insights on how adverse social exposome (ASE) may impact cardiometabolic and brain health. Using complementary metabolomics platforms, we measured blood metabolome as readouts on net influences of exposome factors. Participants from six Alzheimers disease research centers (n=449) were studied with generalizability confirmed in the UK Biobank using its harmonizable metric for ASE (n=380,943). Our results suggest that participants living in ASE have metabolic features often shown to predispose individuals to higher risks for cardiovascular diseases and cognitive decline, with impaired mitochondrial energetics, amino acid and lipid metabolism. Diet, microbiome and chemical exposures may contribute to these metabolic features. Molecular insights from metabolic signatures for ASE allows us to map potential modifiable risk factors that can impact and sustain health including brain health.

An age-related decline in vasodilation mediated by neurovascular nitric oxide (NO) and vasoconstriction driven by the Piezo1 receptor precede the aggregation of soluble brain proteins such as amyloid-{beta} (A{beta}), which then causes further disruption of brain solute homeostasis, ultimately leading to neurodegeneration in Alzheimers disease. Preclinical studies show that restoring these vascular functions increases neurofluidic efflux and improves cognitive outcomes. Here, we tested effects of sublingual NO and/or Piezo1 receptor-targeted mechanotransductive whole-body vibrations (WBVp) in healthy adults (n = 29) on brain fluid dynamics and CNS-to-blood protein efflux using multimodal neuroimaging and blood biomarker analysis. The combined vasomechanic interventions (NO+WBVp) produced a synergistic enhancement of brain fluid transport and markedly increased the efflux of soluble brain-derived proteins (A{beta}40&42, glial fibrillary acidic protein) into the bloodstream. The effects increase with age and the magnitude of NO-induced hypotension. Importantly, the combined intervention was well-tolerated, with no severe adverse physiological responses. Results demonstrate that a simple, non-invasive vasomechanic intervention can transiently promote brain-to-blood protein clearance in humans, highlighting a potentially safe and accessible therapeutic avenue for neurodegenerative conditions characterized by impaired brain solute removal and protein aggregation.

Plant-based dietary strategies may offer a tractable approach to mitigating microbiome disruption and improving outcomes in patients undergoing autologous hematopoietic cell transplantation (auto-HCT) for multiple myeloma, a population in whom intestinal dysbiosis has been linked to infectious complications and inferior survival. We conducted a single-arm study to test the feasibility and biological activity of a high-fiber, plant-based, whole-food meal delivery intervention during the peri-transplant period. Adults with multiple myeloma (n = 22) received fully prepared, plant-based meals for 5 weeks spanning conditioning, neutropenia, and early recovery, with the goal of supporting consumption of nutrient-dense, high-fiber foods despite transplant-related symptoms that often limit oral intake. The primary endpoints were feasibility and tolerability, defined by successful enrollment, adherence to study procedures, and patient-reported intake of study meals; diet was quantified using prospective food diaries and 24-hour dietary recall surveys. Secondary endpoints included changes in gut microbiome composition and function assessed by shotgun metagenomic sequencing and stool short-chain fatty acid (SCFA) measurements. The intervention was feasible and generally well tolerated, with all participants consuming at least some proportion of delivered meals and with adherence sufficient to support planned dietary and correlative analyses. Greater intake of study meals was associated with more pronounced shifts in gut microbial communities, including enrichment of SCFA-producing taxa and compositional changes consistent with a fiber-responsive microbiome. Stool SCFA concentrations increased from baseline to the end of the intervention, suggesting a functional impact of the dietary strategy on microbial metabolite production during the peri-transplant period. These findings demonstrate that a plant-based meal delivery intervention is implementable during auto-HCT and suggest dose-dependent modulation of the gut microbiome and its metabolic output. Larger randomized trials are warranted to determine whether microbiome-targeted nutrition can reduce transplant-related toxicities, enhance immune recovery, and improve disease control in multiple myeloma. The trial is registered at ClinicalTrials.gov (NCT06559709).