Nature Biomedical Engineering

Most routine clinical markers are interpreted using population-based reference intervals, despite being regulated around patient-specific homeostatic setpoints. This mismatch obscures physiologic shifts, inhibiting detection of early disease signatures. Here, we develop a novel Bayesian inference method that adaptively constructs personalized reference intervals using each patients existing health records. In analysis of >100 million lab tests in >800,000 patients, these personalized intervals can be accurately constructed with only minimal prior data, meaning this method can be applied near universally. We show that across 43 common lab markers, patient setpoints are strongly associated with future morbidity, with signal strength increasing as more test data is collected. Deviation from personalized reference intervals provides strong and novel risk signatures across diverse disease states, including hypothyroidism, hematologic cancers, kidney disease, and pregnancy complications. Importantly, personalized reference intervals capture a different risk signature to existing population-based approaches, with the highest risk patients being those who deviate from both intervals simultaneously. In a targeted clinical use case study of iron infusion, use of personalized reference intervals greatly improved prediction of treatment efficacy and allowed precise tracking of treatment responses. Our results illustrate how existing health records can be used to construct personalized benchmarks for nearly all common clinical tests, driving a new paradigm for precision laboratory medicine.

Vocal biomarkers, encompassing voice and speech, have largely been developed for individual conditions in isolation, limiting their generalizability across diseases and recording settings. To address this, we introduce VoiceFM, a contrastive model that learns general-purpose clinical voice representations by aligning audio embeddings with rich clinical metadata. Using the Bridge2AI-Voice dataset (984 primarily English-speaking adult participants, 846 used for training and 138 held out as a temporally separated validation cohort, 40,056 recordings totaling 176 hours across 5 academic medical centers), VoiceFM pairs a fine-tuned Whisper large-v2 encoder with a tabular transformer over 44 clinical features via symmetric InfoNCE loss. Linear probes on frozen VoiceFM embeddings achieve mean AUROC 0.952 +/- 0.005 across five evaluation tasks (control vs disease screening plus four disease categories), significantly outperforming Frozen Whisper (0.926 +/- 0.013, p = 0.013), Frozen HuBERT (0.885 +/- 0.017, p = 0.0009), and the contrastively trained VoiceFM-HuBERT (0.938 +/- 0.006, p = 0.012). On the 138-participant held-out cohort, VoiceFM-Whisper achieves AUROCs of 0.99 for Alzheimer's/dementia/MCI and 0.89 for airway stenosis, demonstrating that the learned representations generalize to participants the model has never seen. VoiceFM representations transfer to three external datasets without retraining and improve few-shot classification. Recording task attribution identifies a small set of speech tasks that match or exceed the full battery's performance, suggesting shorter screening protocols are feasible. Trained predominantly on English audio, VoiceFM transfers without fine-tuning to Spanish-language Parkinson's disease (PD) detection (NeuroVoz, 107 participants, AUROC 0.93 +/- 0.02), with the signal dominated by articulatory rather than phonatory features. A fine-tuned classifier achieves participant-level AUROC 0.87 (sustained 0.85, countdown 0.80) on the mPower smartphone study (585 held-out participants). Together, these results show that contrastive alignment between voice and rich clinical metadata can serve as the basis for a clinical voice foundation model, producing a single set of transferable representations that generalize across diseases, languages, recording conditions, and patients enrolled after model freeze.

Glaucoma phenotyping remains challenging due to disease heterogeneity and single-modality limitations. We introduce a visual field (VF)-guided entangled learning framework that integrates structural and functional signals during training to learn functionally informed macular retinal nerve fiber layer (mRNFL) representations while enabling OCT-only inference. In 5,372 paired MEEI examinations, VF-guided phenotyping identified 9 clinically distinct mRNFL phenotypes with divergent progression rates (MD slopes -0.2 to -1.8 dB/year, P <0.001), improving clustering over OCT-only by 22% (FCM) and 11% (GMM). External evaluation in 74,077 UK Biobank images confirmed generalizability, with improved risk association (r=-0.33 vs r=0.04). Genetic analyses identified 12 additional glaucoma loci compared with OCT-only phenotyping. VF-guided entangled learning improves clinically and genetically coherent mRNFL phenotyping with broad applicability to multimodal medical imaging.

Although GLP1/GIP receptor agonists demonstrate unprecedented weight loss efficacy, their rapid clinical adoption has revealed significant real-world tolerability challenges. To evaluate their dynamic safety profiles, we developed a macro to micro pharmacovigilance framework by combining global FAERS reports with local UT Physician EHR. Macroscopically, we distilled 17 shared adverse events across the drug class from FAERS with disproportionality analysis. Microscopically, local EHR data (289,655 longitudinal treatment sessions across 71,316 patients) revealed 51.6% of GLP1 sessions terminated within 90 days. Furthermore, temporal stratified logistic regression demonstrated that initial exposure (0 to 30 days) correlated strongly with nausea and vomiting, which attenuated in extended sessions, whereas extended exposure (>2 years) uncovered late onset risks, notably incident hepatic steatosis. Ultimately, this time aware framework reveals that GLP1 safety profiles are profoundly duration dependent, providing critical insights into both acute intolerances and long-term medication safety.

Neurosurgical and neuromodulation therapies such as deep brain stimulation (DBS) require millimeter-level accuracy to effectively target functional brain regions. Yet, many neuroanatomical boundaries remain invisible to current imaging and electrophysiology methods, limiting precision and contributing to suboptimal patient outcomes. Here, we introduce a self-supervised artificial intelligence (AI) framework that learns to delineate functional subregions directly from the spectral content of intraoperative local field potential (LFP) recordings, without the need for predefined biomarkers or anatomical labels. The framework identifies physiologic structure across the full spectrum of the signal and, through explainable AI (XAI), reveals the specific frequency components underlying these distinctions. Validated in the subthalamic nucleus (STN), the model aligned with clinically defined borders and rediscovered known beta oscillations. Applied to the motor thalamus in tremor patients, it consistently identified functional transitions corresponding to the ventral oralis posterior (Vop) and ventral intermediate (Vim) nuclei--regions where conventional methods fail to provide reliable boundaries. To assess clinical relevance, physiologically defined clusters were functionally evaluated using monopolar review data at their first DBS clinic postoperative visit, demonstrating distinct stimulation-response profiles across clusters and linking electrophysiologic segmentation to clinically meaningful programming outcomes. These findings demonstrate that intraoperative LFP recordings can be transformed into both a real-time guidance resource and a data-driven platform for biomarker discovery, establishing a foundation for more precise, individualized neuromodulation therapies and advancing our understanding of functional brain organization.

New implantable-electrode fabrication strategies make dense, ultrafine electrode arrays with lower tissue burden increasingly feasible, shifting a key bottleneck for scalable brain-computer interfaces from electrode placement to readout capacity. Electrode pooling, in which multiple electrodes share a readout channel, could relax this bottleneck by combining extracellular signals before acquisition, but it has remained unclear whether such compression preserves the neural population structure needed for behavioral decoding. Here we evaluate this question using software-emulated electrode pooling in mouse sensorimotor cortex during a cue-guided reach-and-grasp task using a high-density microwire array coupled to a CMOS microelectrode-array platform. Pooled recordings retain forelimb kinematic information more effectively than a channel-matched control that discards electrodes. Pooling reduces the separability of electrode-specific spikes and sorted units, indicating a loss of some neuronal detail, but the mixed signals still preserve task-aligned low-dimensional latent dynamics that support decoding. When readout capacity is fixed, this trade-off allows broader electrode coverage to contribute to behaviorally informative population sampling. Together, these results define electrode pooling as a design trade-off for scalable readout, in which some electrode-specific neuronal information is lost but the population dynamics needed for movement decoding remain accessible.

Tau protein aggregation in the brain is a hallmark of Alzheimers disease (AD). Positron emission tomography (PET) is the only in vivo method to visualize tau pathology and estimate both its burden and regional distribution, but the use of tau-PET is constrained by high cost and limited accessibility. Here, we develop a deep learning model to synthesize tau-PET scans from more accessible data: structural magnetic resonance imaging (MRI), demographics, and when available, blood biomarkers. We included 5,191 participants across the AD continuum or with another neurological disorder from 13 cohorts (mean age 70 years, 51% female) and optimized a 3D U-Net neural network with residual and attention units for this task. In held-out test data, synthetic tau-PET reliably modeled tau burden, with correlations of R=0.77-0.86 with true tau-PET across individuals in common AD regions of interest. Spatial similarity between synthetic and true tau-PET was likewise high, with mean regional correlation of R=0.75. Synthetic scans also captured clinically meaningful prognostic information comparable to true tau-PET, including distinction between early (HR=12, p<0.001) and late (HR=45, p<0.001) stages of tau accumulation. These findings demonstrate that clinically informative synthetic tau-PET scans can be generated from widely available modalities using deep learning, potentially offering a scalable and cost-effective approach for estimating tau AD pathology in the brain.

Medical imaging models depend on large, shareable datasets, yet privacy constraints limit data dissemination. Current text-conditioned diffusion models fail to preserve subtle, distributed clinical signals, such as continuous physiological biomarkers, rendering synthetic data insufficient for robust downstream physiological modeling. Here, we evaluate image-to-image (I2I) diffusion as a tunable, privacy-preserving transformation that produces a synthetic counterpart of real images while preserving downstream-relevant information. We fine-tune Stable Diffusion with low-rank adapters on retinal fundus photographs and chest radiographs, assessing fidelity, clinical signal preservation, cross-site transfer, and empirical re-identification risk. I2I consistently outperforms text-to-image generation in image fidelity and in preserving biomarker information. In cross-cohort transfer to an external retinal dataset from the UK Biobank, pretraining on I2I synthetic data performs comparably to real-image pretraining and surpasses it in the smallest fine-tuning sets. Varying I2I strength reveals that the privacy-utility tradeoff is highly modality-dependent: while retinal images achieve practical de-identification, chest X-rays exhibit structural combinatorics that leave them substantially re-identifiable even at high noise strengths, exposing critical boundaries for diffusion-based anonymization. These results position image-conditioned diffusion as a practical approach for generating shareable medical images with tunable de-identification.

Understanding processes such as blood-brain barrier (BBB) disruption and tumor progression can greatly benefit from simultaneous molecular, functional, and hemodynamic imaging in deep tissue, yet few existing imaging modalities can provide all three in a single system. Here, we present an integrated imaging platform that combines 3D photoacoustic tomography with ultrasound localization microscopy (3D-PAULM) to enable intrinsically co-registered, multiparametric imaging. 3D-PAULM unifies multispectral photoacoustic molecular imaging, ultrasound B-mode imaging, microbubble-enhanced power Doppler, and ultrasound localization microscopy, and concurrently measures blood oxygenation, blood perfusion, microvascular flow dynamics, and molecular probes from near-infrared dyes and photoswitchable phytochromes. We apply 3D-PAULM to quantify BBB leakage in focal ischemia and systemic inflammation, and to perform high-sensitivity molecular imaging of solid tumors alongside functional mapping of tumor hypoxia and super-resolved vascular remodeling. Together, these results establish 3D-PAULM as a versatile platform for integrated functional and molecular imaging in deep tissue.

Achieving a trade-off between biological utility and patient privacy remains a key challenge for secure data sharing when applying transcriptomic clinical datasets to artificial intelligence in precision oncology. Here, we introduce the first benchmarking study tailored to high-dimensional clinical transcriptomic cancer data, comparing synthetic data generation methods across three clinical cancer trials. Our framework, SynOmicsBench, combines standardized preprocessing with multidimensional evaluation, prioritizing downstream biological validation alongside statistical fidelity and attack-based privacy assessment. Results indicate that no single method dominated all dimensions, with Gaussian Copula achieving the most balanced performance, followed by Avatar, demonstrating that metric-based similarity alone is insufficient to ensure preservation of higher-order molecular dependencies. Synthetic data consistently reproduced biomedical signal directionality but with attenuated effect sizes and inter-replicate variability, supporting hypothesis generation when multi-seed synthesis is adopted. Collectively, this framework provides a reproducible decision-support tool for method selection and promotes biologically informed, privacy-aware adoption of synthetic data in precision oncology.

Complex human diseases exhibit substantial clinical heterogeneity driven by poorly understood molecular mechanisms, while many also lack sufficient molecular and omics data for mechanistic investigation, hindering therapeutic development. We introduce PiMInfer, a phenotype-to-mechanism framework that leveraged largely available real-world clinical data-based deep phenotypic characterizations with a biomedical knowledge graph approach to resolve disease clinical heterogeneity into phenotype-informed molecular modules, thereby accelerating therapeutic target discovery. We applied PiMInfer to investigate Hidradenitis Suppurativa (HS), an autoimmune skin disease with poorly understood pathogenesis and limited treatment options. PiMInfer identified a coherent, phenotype-informed HS gene module (PiHSM) and functional endotypes, which were validated using multimodal evidence. In silico drug repurposing using PiHSM prioritized Carfilzomib, targeting the immunoproteasome subunit PSMB9, essential for MHC Class I antigen presentation. Preclinical testing using human patient lesional skin explants confirmed its anti-inflammatory activity and demonstrated a significant downregulation of IFN-{gamma}, IL-17, and mTOR signaling pathways within HS lesional microenvironment through single-cell RNA sequencing. PiHSM-based network predictions further suggest a potential enhanced efficacy of combining Carfilzomib with approved HS agents. Collectively, PiMInfer provides a scalable framework that bridges real-world phenome-wide comorbid associations to mechanism-anchored therapeutic discovery, enabling a paradigm shift in precision medicine approaches for complex diseases with limited molecular characterization and in need of better therapeutic strategies.

Noninvasive monitoring of plaque inflammatory dynamics remains an unmet need. We previously developed a monocyte-mimetic nanoprobe, termed MoNP-SPION, for MRI detection of atherosclerotic lesions. Here we demonstrate MoNP-SPION enables longitudinal tracking of plaque inflammatory status in a clinically relevant mouse model. Following 16 weeks of plaque induction, mice were maintained on high-fat diet or switched to chow for 6 weeks to model persistent versus resolving plaque inflammation. MoNP-SPION-enhanced MRI was performed at 3- and 6-weeks post-adjustment, and arterial tissue was collected for histological assessment. Mice maintained on high-fat diet exhibited persistent hypointense T2* signal at the carotid bifurcation and aortic root, whereas chow-transitioned mice showed progressive signal attenuation, consistent with histological evidence of reduced plaque burden and inflammation. These findings establish MoNP-SPION as an effective molecular MRI probe for longitudinal assessment of plaque inflammatory dynamics, supporting its potential for monitoring atherosclerosis progression and therapeutic response.

Colorectal cancer (CRC) is one of the leading causes of cancer-related mortality in men and women. Timely detection and diagnosis are key to management of CRC, which is under-diagnosed because colorectal aberrant crypt foci, hyperplastic polyps, and microadenomas are often missed with conventional colonoscopy. The enzyme cyclooxygenase-2 (COX-2) is overexpressed in early stages of colorectal carcinogenesis and plays an important regulatory role in the process, suggesting that it could be a valuable target for enhanced imaging of nascent disease. Thus, we have generated an alpaca-derived library of 73 COX-2-specific nanobody clones. Here, we describe one such nanobody, F9-K45Q-K77Q-ROX, in which two native lysine residues have been mutated followed by conjugation to a fluorophore at the N-terminus with retention of COX-2-selective binding. The site of fluorophore conjugation and COX-2 binding affinity of F9-K45Q-K77Q-ROX were determined by proteomic and microscale thermophoretic analyses, respectively. In cell culture studies using 1483 human head and neck squamous cell carcinoma cells, F9-K45Q-K77Q-ROX accumulated inside cells and bound to intracellular COX-2, as visualized by fluorescence microscopy. In vivo pharmacokinetic, and toxicological analyses revealed that F9-K45Q-K77Q-ROX is detectable in circulation with a plasma half-life of 17.9 min and there is no short-term toxicity associated with single injections of 10 mg/kg, 20 mg/kg, or 40 mg/kg doses at 24 h post-administration. Noninvasive in vivo fluorescence endoscopic imaging validated tumor-specific accumulation of F9-K45Q-K77Q-ROX in azoxymethane/dextran sodium sulfate-induced colorectal adenomas in mice. This work demonstrates the first COX-2-targeted nanobodies including a fluorescent derivative that offers significant promise for targeted endoscopic imaging of COX-2-expressing neoplasms. Significance StatementCurrent colorectal cancer screening procedures, such as white-light colonoscopy, chromoendoscopy, and narrow-band imaging aim to detect solid colon tumors and precursor lesions. However, these methods tend to detect only raised solid tumors and mature cancers, whereas precursor lesions, such as aberrant crypt foci, hyperplastic polyps, and small adenomas are frequently missed. To address the need for better visualization of early lesions, we developed a library of alpaca-derived nanobodies targeted to cyclooxygenase-2 (COX-2), an enzyme that is overexpressed in colorectal adenomas. COX-2-targeted nanobodies bearing a fluorescent tag accumulate and are retained in colonic adenomas, facilitating their endoscopic visualization. This novel COX-2-targeted nanobody platform may also be valuable for early detection of other neoplastic diseases in which COX-2 overexpression occurs. (Word counts 119, limit 120)

Islet transplantation can restore glycemic control in type 1 diabetes, yet the heterogeneity of patient immune responses and transplant outcomes motivates the need for technologies to monitor the graft. Since transplanted islets are not readily accessible for biopsy due to their diffuse engraftment within the liver, clinical monitoring relies on measurements such as islet mass, blood glucose, and C-peptide levels, which are lagging indicators that change only after substantial graft injury. Here, we developed a minimally invasive synthetic immunological niche (IN) that captures graft-associated immune responses through serial subcutaneous biopsy. We evaluated the IN across murine syngeneic, allogeneic, and autoimmune islet transplant models, including CD40/CD154 costimulatory blockade with anti-CD40L. In syngeneic versus allogeneic recipients, IN identified immune populations and transcriptomic signatures that mirrored the graft and distinguished healthy from rejecting grafts. In anti-CD40L treated allografts, IN revealed innate macrophage- and dendritic cell-associated programs linked to graft acceptance versus rejection, whereas IN from untreated allografts showed stronger adaptive immune signatures. Longitudinal IN profiling further detected progressive inflammatory activation in accepted allografts, indicating persistent subclinical risk. Finally, in an autoimmune allograft model treated with anti-CD40L plus rapamycin, IN identified a 13-gene signature that separated early from late rejection trajectories and distinguished autoimmune-from alloimmune-associated rejection programs. Overall, these findings establish IN as a surrogate tissue for minimally invasive monitoring of islet graft and early detection of rejection-associated immune dysregulation. One Sentence SummaryAn engineered immunological niche captures distinct immune signatures of allo- and auto-mediated islet transplant rejection

Three-dimensional stem cell-derived neural organoids provide a promising platform for investigating early brain development and interregional circuit formation. Although co-culture of region-specific organoids into assembloids has enabled the study of cortical and subcortical interactions, these models lack directional specificity and spatial control, limiting their ability to recapitulate canonical circuit architecture. Here, we present a microfluidic platform for constructing directional and tunable interregional circuits while preserving anatomical distinction. This system, which we term "directoids" incorporates micropatterned polydimethylsiloxane (PDMS) microstructures to control uni- and bidirectional axonal growth between cortical and thalamic organoids. We observed a 70.4% success rate of axons traversing the full channel length in the permissive direction and reaching the opposing organoid, whereas no neurites successfully crossed the probative direction. These results demonstrate robust directionally bias in axon outgrowth and establish a scalable, reproducible strategy for controlling asymmetric connectivity between anatomically distinct neural organoids. Using high-density CMOS microelectrode arrays, we further validated directional tuning of extracellular action potential propagation within directoid microchannels, a feature not observed in straight-channel connectoid controls. Directoids also exhibited significant asymmetry in firing rates between channel entry and exit sites, consistent with engineered bias in signal flow. This provides an experimental paradigm for dissecting how anatomical connectivity and functional activity converge to shape neuronal networks. Together, these findings establish a microfluidic platform for investigating the mechanisms underlying hierarchical circuit formation, regional specification, and functional integration in developing human neural organoid models at cellular resolution not possible in vivo.

Transmissible hospital-acquired infections (HAIs) arise from complex, time-varying interactions among patients, healthcare workers, and clinical environments. Although data-driven approaches like graph neural networks (GNNs) effectively model these contacts, they often function as black boxes that over-look established epidemiological principles, limiting interpretability and clinical trust. Inspired by physics-informed neural networks, we propose a epidemiology-informed GNN (EIGNN) framework for patient-level state transitions prediction in dynamic hospital settings, integrating mechanistic epidemiological models into GNNs in a principled manner. Patient-level risk factors learned from dynamic contact networks are jointly leveraged to infer latent epidemiological states, predict state transitions across multiple horizons, and estimate key epidemiological parameters, including transmission and recovery rates. We evaluate the approach on a real-world hospital-onset COVID-19 cohort and two public datasets simulating viral and bacterial HAIs. Across multiple architectures and horizons, EIGNNs achieves AUC-ROC up to 98.46% while providing interpretable, mechanistically consistent insights, offering a transparent tool for infection prevention and control.

Light is a modifiable determinant of health, yet real-world exposure assessment is often reduced to illuminance alone, lacks environmental context, or relies on privacy-sensitive sensing. We present SpectraVita, a low-cost, compact multispectral wearable that continuously samples 11 ultraviolet-to-near-infrared bands and, through a privacy-preserving pipeline without cameras or location tracking, produces interpretable digital phenotypes of lighting environment (natural vs. artificial and source type) and vegetation context alongside standard visual and non-visual light metrics. In extensive in-the-wild recordings spanning diverse scenes, times of day, weather conditions, and light sources, we observe distinctive spectral signatures that enable supervised models to achieve a macro-averaged F1 score of 0.988{+/-}0.004 for light-source classification and green-space detection in boundary-free environments. A sensor-derived normalized difference vegetation index (NDVI) emerges as an explainable, physically grounded marker linking natural light exposure and greenness. Robustness is supported by scenario-shift testing, image-segmentation validation, and mixed-environment experiments that demonstrate sensitivity to partial and transient exposures, as well as by longitudinal stationary monitoring and deployment in a cohort of thousands of participants capturing seasonal and behavioral variability. SpectraVita enables individualized, privacy-preserving, longitudinal monitoring of light and greenness exposure at scale, addressing a key measurement gap for precision and population health studies of daily photic environments.

The limiting step in immune-active drug development is increasingly not candidate generation but testing whether a candidate therapy is effective in a system that preserves tissue architecture, vascular exposure, multicellular interaction, and repeated pharmacodynamic sampling without patient exposure. We developed an acellular normothermic machine-perfusion platform for intact porcine spleen designed as a translational immune-organ assay. Across independent acellular perfusions, the circuit maintained physiologic parameters, preserved red- and white-pulp histology, and yielded viable effluent cells suitable for serial flow cytometry and multiomics. High-dose methylprednisolone was used as a clinically familiar perturbation to determine whether the platform could resolve steroid immunosuppression at mechanistically distinct levels. Effluent RNA-seq identified canonical glucocorticoid-responsive transcriptional programs, including DUSP1, FKBP5, PER1, DDIT4, SGK1, KLF9, ANXA1, NF-{kappa}B feedback regulators, and JAK/STAT suppressor pathways. SOCS3 was a prominent early signal in the perfusion transcriptome and was validated orthogonally at the protein level in prednisone-treated, CD3/CD28-activated primary murine splenocytes, strengthening its role as a candidate pharmacodynamic marker. In parallel, data independent acquisition (DIA) proteomics of effluent cell pellets nominated a non-transcriptional protein-level response: a Sus scrofa LGALS13-annotated, CLC/Galectin-10-like galectin detected despite absence of the corresponding effluent-cell transcript. Because this porcine LGALS13-annotated protein group is treated here as an orthologous CLC/Galectin-10-like signal rather than as canonical human placental Galectin-13/PP13, we tested recombinant human Galectin-10 in vitro. Human Galectin-10 induced marked apoptosis of CD3/CD28-stimulated Jurkat cells, prioritizing this axis for future mechanistic testing without proving causality in the perfused spleen. These data establish acellular spleen perfusion as a serial multiomic platform for translational immunopharmacology and motivate deployment with otherwise-discarded human donor spleens. One sentence summaryAn acellular intact-spleen perfusion platform enables serial cellular, transcriptomic, proteomic, and functional pharmacodynamic sampling that identifies steroid-responsive transcriptional programs, validates SOCS3 protein induction, and nominates a CLC/Galectin-10-like non-transcriptional immunosuppressive axis for translation to discarded human donor spleens.

The development of clinically effective CAR-T cell therapies for solid tumors requires careful optimization of receptor design, functional fitness, and manufacturability. While advancing low-affinity HER2-targeting CAR-T cells toward clinical application, we found that the candidate with the strongest in vivo antitumor activity--comprising a CD8 hinge and transmembrane region and a 4-1BB co-stimulatory domain--exhibited measurable tonic signaling. This basal antigen-independent signaling, likely driven by high CAR surface expression, was associated with increased apoptosis and reduced ex vivo expansion under research-grade manufacturing conditions. Modification of the transmembrane domain reduced CAR surface expression but did not alleviate tonic signaling and instead impaired antitumor activity. By contrast, transient pharmacologic inhibition of CAR signaling with dasatinib rescued expansion and reduced apoptosis in small-scale research cultures. Notably, these tonic-signaling-associated defects were largely absent during large-scale, GMP-compliant manufacturing, which enabled robust CAR-T cell expansion without additional benefit from dasatinib supplementation. Together, these findings show that tonic signaling is not inherently detrimental to CAR-T cell performance and that its functional consequences are highly dependent on manufacturing context. Our study underscores the importance of evaluating CAR candidates within clinically relevant production platforms and supports the advancement of this 4-1BB-based HER2-specific CAR-T cell product toward clinical testing.

Anti-amyloid immunotherapies have recently demonstrated the first significant slowing of cognitive decline in Alzheimers disease (AD), yet clinical benefit varies markedly across drugs and scales with the completeness of amyloid clearance. Pharmacokinetic/pharmacodynamic (PK/PD) models are currently the standard tool for trial simulation, but they typically operate on single biomarkers and rely on drug-concentration assumptions, leaving the multi-scale cascade from amyloid clearance through tau, neurodegeneration, and cognition largely unmodelled. No existing framework has been jointly validated against the quantitative outcomes of multiple real-world phase III trials, spanning clearance kinetics, multi-modal biomarker trajectories, and statistical power. We present a trial simulation platform based on SimulAD, a disease progression model trained exclusively on longitudinal observational data from ADNI, with no access to trial-arm labels or drug-specific outcomes. SimulAD encodes intervention as piecewise amyloid clearance terms within a latent ordinary differential equation system that jointly governs amyloid, tau, structural MRI, and cognitive trajectories under the amyloid cascade hypothesis. We retrospectively simulated six landmark phase III anti-amyloid trials (TRAILBLAZER-ALZ2, CLARITY AD, EMERGE and ENGAGE, GRADUATE I and GRADUATE II) using a single trained model with trial-specific calibration limited to amyloid clearance kinetics. SimulAD reproduced published mean centiloid reductions within 5% error across all six trials and generated CDR-SB distributions broadly consistent with reported placebo and treated-arm outcomes. In a retrospective power analysis, calibrated simulations separated the three positive from the three null trials, with EMERGE near the decision boundary and ENGAGE and both GRADUATE trials below it. Across trials, higher amyloid-clearance rates were associated with larger calibrated clinical effects and lower estimated sample sizes. These results establish SimulAD as a valid disease-progression-centric trial simulator providing quantitative guidance on sample size planning and treatment kinetics optimisation that is grounded in the full multi-modal biomarker cascade of AD.