Med

BackgroundWhole-genome sequencing (WGS) has improved the diagnosis of rare genetic disorders, yet interpretation of non-coding variants that affect splicing remains challenging. In silico predictions alone are insufficient, and short-read RNA sequencing may fail to capture complex or low-abundance splicing events. Targeted amplicon-based long-read RNA sequencing (Amp-LRS) offers a cost-effective approach for functional validation of candidate splice-altering variants. MethodsWe applied Amp-LRS to five patients with neurological disorders (central nervous system, peripheral nervous system, or muscle) harbouring candidate non-coding variants predicted to alter splicing. RNA was extracted from fibroblasts or peripheral blood, and full-length transcript amplicons were sequenced using Oxford Nanopore Technologies. Nonsense-mediated decay (NMD) inhibition was performed on fibroblast cultures using cycloheximide. ResultsAmp-LRS validated all five candidate variants, including intronic and UTR variants in POLR3A, OPA1, PYROXD1, GDAP1, and SPG11. Aberrant splicing events included exon skipping, intron retention, cryptic splice site activation, and pseudoexon inclusion, often resulting in frameshifts and premature termination codons. For POLR3A and OPA1, multiple abnormal isoforms arose from single variants, highlighting the complexity of splicing disruption. Some pathogenic effects were detectable only in a minority of reads and variably enriched by NMD inhibition, consistent with being hypomorphic. The approach was successfully applied using accessible tissues and enabled multiplexed sequencing at low per-sample cost. ConclusionsAmp-LRS is a sensitive, versatile, and cost-effective method for functional assessment of non-coding splice-altering variants identified by WGS. By enabling full-length transcript analysis from accessible tissues, this approach improves interpretation of variants of uncertain significance and could enhance molecular diagnosis in rare neurological diseases.

Vitamin B12 is critical for hematopoiesis and myelination.1 Deficiency can cause neurologic deficits including loss of coordination, spasticity, and cognitive decline.2,3,4 However, diagnosis relies on vitamin B12 measurement in the blood which may not accurately reflect levels in the brain. Here, we discovered an autoimmune cause of vitamin B12 deficiency restricted to the central nervous system (CNS), termed autoimmune B12 central deficiency (ABCD). Using programmable phage display, we identified an autoantibody targeting the transcobalamin receptor (CD320) in a patient with progressive tremor, ataxia, and scanning speech. Patient immunoglobulins impaired cellular uptake of vitamin B12 in vitro. Despite normal serum levels, vitamin B12 was nearly undetectable in her cerebrospinal fluid (CSF). Immunosuppressive treatment and high-dose systemic vitamin B12 supplementation were associated with increased CSF B12 levels and clinical improvement. Autoantibodies targeting the same epitope of CD320 were identified in 7 other patients with neurologic deficits of unknown etiology and in 6 percent of healthy controls. In 132 paired serum and CSF samples, detection of anti-CD320 in the blood predicted B12 deficiency in the brain. These findings elucidate a new autoimmune cause of metabolic neurologic disease that may be amenable to immunomodulatory treatment and/or nutritional supplementation.

After one year of the COVID-19 pandemic, over 130 million individuals worldwide have been infected with the novel coronavirus, yet the post-acute sequelae of COVID-19 (PASC), also referred to as the long COVID syndrome, remains mostly uncharacterized. We leveraged machine-augmented curation of the physician notes from electronic health records (EHRs) across the multi-state Mayo Clinic health system to retrospectively contrast the occurrence of symptoms and diseases in COVID-19 patients in the post-COVID period relative to the pre-COVID period (n=6,413). Through comparison of the frequency of 10,039 signs and symptoms before and after diagnosis, we identified an increase in hypertensive chronic kidney disease (OR 47.3, 95% CI 23.9-93.6, p=3.50x10-9), thromboembolism (OR 3.84, 95% CI 3.22-4.57, p=1.18x10-4), and hair loss (OR 2.44, 95% CI 2.15-2.76, p=8.46x10-3) in COVID-19 patients three to six months after diagnosis. The sequelae associated with long COVID were notably different among male vs female patients and patients above vs under 55 years old, with the hair loss enrichment found primarily in females and the thromboembolism enrichment in males. These findings compel targeted investigations into what may be persistent dermatologic, cardiovascular, and coagulopathic phenotypes following SARS-CoV-2 infection.

The automatic extraction of medication mentions from social media data is critical for pharmacovigilance and public health monitoring. In this study, we present an end-to-end generative approach based on instruction-tuned large language models (LLMs) for medication mention extraction from Twitter. Reformulating the task as a text-to-text generation problem, our models achieve state-of-the-art results on both fine-grained span extraction and coarse-grained tweet-level classification, surpassing traditional sequence labeling baselines and previous best-performing systems. We demonstrate that fine-tuning Flan-T5 models enables efficient and accurate extraction while simplifying the architecture by eliminating complex multi-stage pipelines. Additionally, we show that lexicon-based filtering further improves performance by reducing false positives. Our models are publicly available, providing high-performing and efficient tools for large-scale pharmacological analysis of social media data.

Enterally-administered nimodipine is the only approved drug formulation available in the United States for treatment of patients with aneurysmal subarachnoid hemorrhage. Intravenous nimodipine is available in other countries but it contains a high concentration of ethanol that is irritating to the vasculature, can alter the effects of other medications, impair neurological assessments and is potentially harmful to the liver. We developed a sterile aqueous solution of nimodipine solubilized in polysorbate 80 micelles (GTX-104) that circumvents these problems. GTX-104 has been administered to 168 healthy human volunteers in 2 studies. We report the second study here, a phase 1, single center, randomized, 2-period cross over study that assessed the pharmacokinetics of GTX-104 and oral nimodipine capsules, which is the reference standard, in 58 healthy human volunteers. GTX-104 was administered for 72 hours as a continuous infusion of 0.15 mg/hour with a 30 minute bolus infusion of 4 mg every 4 hours. Nimodipine capsules were administered orally at a dose of 60 mg every 4 hours for 72 hours. The maximum plasma concentrations after the first dose of each formulation were similar (GTX-104: 63 ng/mL, n=57 versus nimodipine capsules: 69 ng/mL, n=56, ratio and 90% confidence interval [CI] of geometric means: 92% [90% CI: 82-104%]). The areas under the concentration-time curves on the 3rd day at steady state also were the same (GTX-104: 497 ng*h/mL, n=55 versus nimodipine capsules: 495 ng*h/mL, n=56, ratio and 90% CI of geometric means: 106% [90% CI: 99-114%]). The secondary pharmacokinetic parameters (daily maximum concentration at steady-state and time to maximum concentration) were also similar for the 2 formulations. The variability in PK parameters was less for GTX-104 compared to oral nimodipine. The average oral bioavailability for nimodipine capsules was 7%. These results enabled a Phase 3 safety study of GTX-104 in humans with aneurysmal subarachnoid hemorrhage.

Noninvasive biomarkers for neurodegenerative diseases are urgently needed for earlier detection, monitoring, and intervention. To address this, we developed a blood-based cfDNA methylation platform. A whole-genome nanopore methylation atlas of six primary human neural cell types--cortical, dopaminergic, and spinal motor neurons, astrocytes, Schwann cells, microglia--was used to train classifiers that assign cfDNA to its neuronal or glial origin. Classifiers were validated in silico and applied to 219 plasma samples from patients with Alzheimers disease (AD), Parkinsons disease (PD), amyotrophic lateral sclerosis (ALS), and controls. Cortical cfDNA was elevated in AD, dopaminergic cfDNA in PD, and spinal motor neuron cfDNA in ALS, with predictive modeling achieving AUCs >0.98. In mild cognitive impairment, cortical cfDNA elevations identified individuals who later progressed to AD, supporting predictive utility. A multivariate model integrating multiple neurons improved accuracy, suggesting cfDNA methylation profiling as a scalable framework for noninvasive neurodegeneration detection and monitoring.

BackgroundThe extracellular microenvironment modulates cancer behavior. Although radiographic contrast enhancement is an ominous finding in gliomas, it remains unclear if the associated blood-brain barrier disruption merely reflects or functionally supports tumor aggressiveness. MethodsWe utilized intra-operative microdialysis to sample the extracellular metabolome of radiographically diverse regions during fifteen neurosurgical resections. The global extracellular metabolome of recovered microdialysate was evaluated via ultra-performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS) and assessed using enrichment and correlation analyses. ResultsAmong 162 named metabolites identified via ultra-performance liquid chromatography tandem mass spectrometry, guanidinoacetate (GAA), was 126.32x higher in enhancing tumor than in adjacent brain. 48 additional metabolites were 2.05-10.18x more abundant in enhancing tumor than brain. With exception of GAA, and 2-HG in IDH-mutant gliomas, differences between non-enhancing tumor and brain microdialysate were comparatively modest and less consistent. The enhancing but not the non-enhancing glioma metabolome was significantly enriched for plasma-associated metabolites largely comprising amino acids and carnitines. ConclusionsOur findings suggest that metabolite diffusion through a disrupted blood-brain barrier may largely define the enhancing extracellular glioma metabolome. Future studies are needed to determine how the altered extracellular metabolome impacts glioma behavior.

Multi-consistency testing during flexible endoscopic evaluation of swallowing (FEES) is clinically necessary but introduces selection bias: worst scores inflate severity because the number of consistencies tested covaries with disease severity. In this retrospective observational study of hospitalized neurological patients, we derived and validated the FEES Dysphagia Index (FDI) in two temporally independent cohorts (Cohort 1: 2013-2018, N=1,257; Cohort 2: 2021-2025, N=1,686) from a single center. FDI-S averages Penetration-Aspiration Scale (PAS) scores across tested consistencies (0-100 scale); FDI-E uses Yale Pharyngeal Residue scores; FDI-C combines both. Selection bias was quantified using sequential branching-tree inverse probability weighting (IPW). Worst PAS overestimated severity by 24%; FDI deviated by <2%. FDI-C was significantly superior to Worst PAS for hospital-acquired pneumonia (HAP; AUC 0.70 vs. 0.60, p<0.001), mortality (0.71 vs. 0.62, p=0.040), and restricted oral intake (0.90 vs. 0.74, p<0.001), and statistically equivalent to clinician-rated severity. FDI-C mapped linearly onto ordinal Functional Oral Intake Scale values (FOIS; proportional odds RCS p=0.99). With functional status and diagnosis, FDI-C reconstructed the clinicians oral intake recommendation with AUC up to 0.93. The FDI-C-mortality relationship was sigmoidal with a clinically relevant transition zone between [~]50 and [~]85. FDI-C is a bias-resilient, bedside-calculable score with interval-scale properties that captures expert clinical judgment, suitable as both a clinical decision support tool and a continuous research endpoint.

Recent studies have investigated post-acute sequelae of SARS-CoV-2 infection (PASC) using real-world patient data such as electronic health records (EHR). Prior studies have typically been conducted on patient cohorts with small sample sizes1 or specific patient populations2,3 limiting generalizability. This study aims to characterize PASC using the EHR data warehouses from two large national patient-centered clinical research networks (PCORnet), INSIGHT and OneFlorida+, which include 11 million patients in New York City (NYC) and 16.8 million patients in Florida respectively. With a high-throughput causal inference pipeline using high-dimensional inverse propensity score adjustment, we identified a broad list of diagnoses and medications with significantly higher incidence 30-180 days after the laboratory-confirmed SARS-CoV-2 infection compared to non-infected patients. We found more PASC diagnoses and a higher risk of PASC in NYC than in Florida, which highlights the heterogeneity of PASC in different populations.

Most clinical large language model (LLM) benchmarks rely on clean, concise vignettes that do not reflect the noisy, long-form documentation typical of real clinical records. How LLM performance degrades under realistic chart conditions remains poorly characterised. Here we test whether structured retrieval workflows protect National Institutes of Health Stroke Scale (NIHSS) scoring accuracy under systematic context stress. Using 100 de-identified acute stroke cases and a fully crossed 4 x 4 x 3 x 3 condition matrix (144 conditions per case), we vary context acquisition method, document length, distractor load and critical-information position across four Gemini models (57,047 retained runs). Structured retrieval reduces mean absolute error (MAE) from 4.58 to 2.96 points relative to non-agentic baselines (mean gain 1.62 MAE points; 95% CI 1.57 to 1.67; 35% relative reduction), with consistent gains across all 36 stress combinations. Lower-cost models show disproportionately larger gains (2.76 versus 0.45 MAE points). Tool-retrieved pipelines outperform retrieval-augmented generation in 33 of 36 combinations. These findings indicate that retrieval architecture, rather than model scale alone, is a tractable lever for robust, equitable clinical LLM deployment.

Integrating structured clinical knowledge into artificial intelligence (AI) models remains a major challenge. Medical codes primarily reflect administrative workflows rather than clinical reason ing, limiting AI models ability to capture true clinical relationships and undermining their gen eralizability. To address this, we introduce ClinGraph, a clinical knowledge graph that integrates eight EHR-based vocabularies, and ClinVec, a set of 153,166 clinical code embeddings derived from ClinGraph using a graph transformer neural network. ClinVec provides a machine-readable representation of clinical knowledge that captures semantic relationships among diagnoses, med ications, laboratory tests, and procedures. Panels of clinicians from multiple institutions evalu ated the embeddings across 96 diseases and more than 3,000 clinical codes, confirming their alignment with expert knowledge. In a retrospective analysis of 4.57 million patients from Clalit Health Services, we show that ClinVec supports phenotype risk scoring and stratifies individuals by survival outcomes. We further demonstrate that injecting ClinVec into large language models improves performance on medical question answering, including for region-specific clinical sce narios. ClinVec enables structured clinical knowledge to be injected into predictive and genera tive AI models, bridging the gap between EHR codes and clinical reasoning.

Despite advances in cancer therapeutics, early detection is often the best prognostic indicator for survival (1). People with Li-Fraumeni syndrome harbor a germline pathogenic variant in the tumor suppressor gene TP53 (2) and face a near 100% lifetime risk of developing a wide spectrum of, often multiple, cancers (3). TP53 mutation carriers routinely undergo intensive surveillance protocols which, although associated with significantly improved survival, are burdensome to both the patient and the health care system (4). Liquid biopsy, the analysis of cell-free DNA fragments in bodily fluids, has become an attractive tool for a range of clinical applications, including early cancer detection, because of its ability to provide real-time holistic insight into the cellular milieu (5). Here, we assess the efficacy of a multi-modal liquid biopsy assay that integrates a targeted gene panel, shallow whole genome, and cell-free methylated DNA immunoprecipitation sequencing for the early detection of cancer in a cohort of Li-Fraumeni syndrome patients: 196 blood samples from 89 patients, of which 26 were pediatric and 63 were adults. Our integrated analysis was able to detect a cancer-associated signal in 79.4% of samples from patients with active cancer, a 37.5% - 58.8% improvement over each individual analysis. Through analysis of patient plasma at cancer negative timepoints, we were able to detect cancer-associated signals up to 16 months prior to occurrence of cancer as detected by conventional clinical modalities in 17.6% of TP53 mutation carriers. This study provides a framework for the integration of liquid biopsy into current surveillance methods for patients with Li-Fraumeni syndrome.

Due to a combination of asymptomatic or undiagnosed infections, the proportion of the United States population infected with SARS-CoV-2 was unclear from the beginning of the pandemic. We previously established a platform to screen for SARS-CoV-2 positivity across a representative proportion of the US population, from which we reported that almost 17 million Americans were estimated to have had undocumented infections in the Spring of 2020. Since then, vaccine rollout and prevalence of different SARS-CoV-2 variants have further altered seropositivity trends within the United States population. To explore the longitudinal impacts of the pandemic and vaccine responses on seropositivity, we re-enrolled participants from our baseline study in a 6- and 12-month follow-up study to develop a longitudinal antibody profile capable of representing seropositivity within the United States during a critical period just prior to and during the initiation of vaccine rollout. Initial measurements showed that, since July 2020, seropositivity elevated within this population from 4.8% at baseline to 36.2% and 89.3% at 6 and 12 months, respectively. We also evaluated nucleocapsid seropositivity and compared to spike seropositivity to identify trends in infection versus vaccination relative to baseline. These data serve as a window into a critical timeframe within the COVID-19 pandemic response and serve as a resource that could be used in subsequent respiratory illness outbreaks.

While the public health burden of SARS-CoV-2 infection has lessened due to natural and vaccine-acquired immunity, the emergence of less virulent variants, and antiviral medications, COVID-19 continues to take a significant toll. There are > 10,000 new hospitalizations per week in the U.S., many of whom develop post-acute sequelae of SARS-CoV-2 (PASC), or "long COVID", with long-term health issues and compromised quality of life. Early identification of individuals at high risk of severe COVID-19 is key for monitoring and supporting respiratory status and improving outcomes. Therefore, precision tools for early detection of patients at high risk of severe disease can reduce morbidity and mortality. Here we report an untargeted and longitudinal metabolomic study of plasma derived from adult patients with COVID-19. One-carbon metabolism, a pathway previously shown as critical for viral propagation and disease progression, and a potential target for COVID-19 treatment, scored strongly as differentially abundant in patients with severe COVID-19. A follow-up targeted metabolite profiling revealed that one arm of the one-carbon metabolism pathway, the methionine cycle, is a major driver of the metabolic profile associated with disease severity. The methionine cycle produces S-adenosylmethionine (SAM), the methyl group donor important for methylation of DNA, RNA, and proteins, and its high abundance was reported to correlate with disease severity. Further, genomic data from the profiled patients revealed a genetic contributor to methionine metabolism and identified the C677T allele of the MTHFR gene as a pre-existing predictor of disease trajectory - patients homozygous for the MTHFR C677T have higher incidence of experiencing severe disease. Our results raise the possibility that screening for the common genetic MTHFR variant may be an actionable approach to stratify risk of COVID severity and may inform novel precision COVID-19 treatment strategies.

The gut microbiome is increasingly implicated in multiple sclerosis (MS), yet findings across studies remain inconsistent due to small sample sizes and methodological variability. We performed the largest MS microbiome meta-analysis to date, integrating 14 datasets (n = 1,493; 777 RRMS, 87 PPMS, 66 SPMS, 563 healthy controls) with standardized preprocessing. We analyzed community diversity, taxonomic composition, co-occurrence networks, and enterotype clustering, accounting for technical and demographic confounders. Alpha diversity did not differ significantly between MS subtypes, treatment groups, or controls. However, beta diversity revealed small, but significant differences between MS patients and controls, and among MS subtypes. Differential abundance analysis identified taxa enriched in MS, Akkermansia, and Eisenbergiella, which were notably associated with progressive disease. A random forest classifier distinguished RRMS from progressive MS with [~]84% accuracy. Co-occurrence networks differed by subtype: RRMS networks were fragmented, SPMS were more cohesive, and PPMS were sparse, but highly modular. Hub taxa composition shifted accordingly. Enterotyping revealed five clusters; RRMS samples were enriched in Akkermansia- and Blautia-dominated types, while controls showed enrichment in Faecalibacterium- and Christensenellaceae-rich clusters. Progressive MS lacked a dominant enterotype. Together, these findings support a "fragmentation hypothesis" in which microbial ecosystems become increasingly unstable with MS progression.

The high heterogeneity of pediatric cancers presents significant diagnostic challenges, underscoring the need for accurate classification. Although molecular profiling supports first-line diagnostics and guides treatment, it can delay final diagnosis. While Nanopore-based methylation analysis has enabled rapid CNS tumor diagnosis, its application to pediatric solid tumors and lymphomas has remained largely unexplored. We developed Tucan, a deep-learning classifier trained on 3,818 methylation array profiles representing 84 subtypes, designed to classify tumors from sparse Nanopore methylation data. In retrospective validation (n=514), Tucan generated confident predictions (CFT[&ge;] 0.7) within 30 minutes of sequencing in 385 cases, achieving 372 correct diagnoses (F1-score: 0.98). In prospective testing (n=74; 63 classifiable), 52 samples reached the confidence threshold with 96% accuracy, confirming the original diagnosis in 47 cases and correctly refining or revising it in three. Together, Tucan enables rapid, high-confidence molecular classification of pediatric solid tumors and lymphomas.

Testing for DNA mismatch repair deficiency (MMRd) is recommended for all colorectal cancers (CRC). Automation of this would facilitate precision medicine, particularly if it provided information on likely aetiology. We developed AIMMer, an AI-based method for determination of MMR protein expression at single cell level in routine pathology samples. We applied it to over 2,000 colorectal cancers (CRC) from the SCOT clinical trial, which compared 3 vs 6 months of oxaliplatin-based adjuvant chemotherapy (FOLFOX or CAPOX). Benchmarking of AIMMeR against pathologist ground truth MMR calls revealed AUROC of 0.98, and positive predictive value (PPV) greater than 95% for the commonest pa[ern of somatic MMRd, and for retained MMR expression. Analysis of CRC recurrence confirmed the prognostic value of MMRd in oxaliplatin-treated patients. While MMRd did not predict differential benefit from chemotherapy duration, it correlated with difference in clinical outcome by chemotherapy regimen (PInteraction=0.04). AIMMeR holds promise to reduce pathologist workflow and streamline clinical diagnostics in CRC.

BackgroundFirefighters have frequent exposure to compounds shown to increase risk of esophageal neoplasia. EsoGuard(R) (EG) is a DNA biomarker assay that can be utilized with efficiency and high tolerability as a triage to endoscopy for diagnosis of patients with Barretts Esophagus (BE), a known precursor to esophageal adenocarcinoma (EAC). This diagnostic tool may facilitate disease testing among busy at-risk firefighters. MethodsRetrospective analysis of prospectively collected clinical utility (CU) data for use of EG as a triage to more invasive endoscopic evaluation. EG was performed on esophageal cell samples collected with the nonendoscopic EsoCheck(R) (EC) device during two large cancer and pre-cancer screening events for firefighters in San Antonio, TX, in January 2023. CU was evaluated by provider impact assessment. Results388 firefighters were identified for EG testing, of which >99% (385/388) successfully completed EC cell collection. Over 96% (372/385) of tests had binary results; the remaining <4% failed analysis due to insufficient DNA. The EG positivity rate was 7.3% (28/385), all of whom were referred for specialist and upper endoscopy evaluation. Among those who tested negative, none were referred for further diagnostic workup. This represented a 100% concordance between EG results and physician management decisions. ConclusionsThis study capturing real-world data on use of EG in a population of firefighters demonstrates its ability to test many individuals rapidly and efficiently in a well-tolerated fashion, and reliable use of the test to triage individuals prior to pursuing more invasive and time-consuming diagnostic approaches.

Fecal microbiota transplantation (FMT) is an effective therapy for recurrent Clostridioides difficile infection (rCDI) but has undefined composition and poor scalability. In vitro manufactured live biotherapeutic products (LBP) enable both scalability and defined strain composition but with higher manufacturing complexity, resulting in few LBP trials. We developed an accessible platform to produce human-grade LBPs. We provide regulatory documentation and manufacturing protocols to facilitate translating microbiome advances to human trials. With this platform, we conduct the first direct comparison of the same bacterial strains administered after in vitro manufacturing (LBP) compared to donor sourced (FMT) across two doses. In a phase 1b trial (n=18), an endoscopic dose of the 15-strain consortium MTC01 was safe with rCDI prevention eight weeks after dosing in seven out of nine LBP patients, similar to eight out of nine FMT patients. Notably, MTC01 strain engraftment was superior to FMT at higher doses.

The majority of mutational signatures have been characterized in tumors from Western countries and the degree to which mutational signatures are similar or different in Eastern populations has not been fully explored. We leveraged a large-scale clinical sequencing cohort of tumors from a Chinese population containing 25 tumor types and found that the highly active mutational signatures were similar to those previously characterized1,2. The aristolochic acid signature SBS22 was observed in four soft tissue sarcomas and the POLE-associated signature SBS10 was observed in a gallbladder carcinoma. In lung adenocarcinoma, the polycyclic aromatic hydrocarbon (PAH) signature SBS4 was significantly higher in males compared to females but not associated with smoking status. The UV-associated signature SBS7 was significantly lower in cutaneous melanomas from the Chinese population compared to a similar American cohort. Overall, these results add to our understanding of the mutational processes that contribute to tumors from the Chinese population.