Biology subjects

Park, J.

Publications and source records attributed to Park, J..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

Unlocking Sensitive Data with SPHERE in the Age of AI

Sensitive human data underpin discoveries across medicine, biology and the social sciences, yet privacy regulation often prevents sharing them with collaborators or artificial intelligence (AI) systems. We introduce SPHERE, a model-free method that makes sensitive datasets directly usable by AI and shareable for open science as a synthetic twin, while the original records never leave the local environment. Across 33 datasets spanning five scientific domains, SPHERE protects individual privacy against adversarial re-identification attacks while preserving the data's statistical structure: means, variances and correlations are reproduced exactly, effect size and P value in linear statistical analysis is numerically identical, nonlinear machine-learning utility is retained, and each twin is generated in seconds on a laptop. Frontier AI agents running on the twin reach the same scientific conclusions as on the original records. Analyses of the twin reproduce genome- and proteome-wide results at UK Biobank scale and recover the findings of landmark studies across three independent cohorts and consortia. The approach also extends to deep-learning embeddings across language, vision and time-series, with minimal utility loss. We make the Stanford Alzheimer's Disease Research Center cohort openly available for the first time, as a SPHERE twin spanning nine modalities that any registered researcher can analyze without an approval process. We release SPHERE with certification of each twin's privacy and fidelity, and an AI agent that autonomously executes research tasks on sensitive data without ever accessing it. Sensitive datasets that are currently closed to research could thus become routine inputs to open science and AI to enable key discoveries.

bioinformatics↗

bioRxiv2026

Rewiring of Integrin Signaling and Cell-cycle Deregulation Drive SMARCB1-Deficient Epithelioid Sarcoma

Epithelioid sarcoma (EPS) is an aggressive soft-tissue sarcoma characterized by loss of the chromatin-remodeling subunit SMARCB1. The oncogenic programs driving EPS remain poorly understood. Through CRISPR loss-of-function screens, we identified conserved dependencies on integrin signaling components and cyclin-dependent kinases (CDKs). Genetic disruption of integrin subunit alpha V (ITGAV)-centered signaling impaired epithelioid cluster formation and reduced MYC expression. SMARCB1 re-expression phenocopied these effects and revealed that SMARCB1 loss selectively represses context-dependent integrin subunits while preserving an ITGAV-centered pro-survival axis, associated with altered BAF complex occupancy. Analysis of EPS cell lines and primary tumors revealed frequent genetic or epigenetic inactivation of CDKN2A/p16, indicating that loss of cell-cycle control is a key cooperating event in EPS development and providing a mechanistic rationale for targeting CDK4/6. Together, these findings establish integrin-driven oncogenic signaling coupled with disruption of cell-cycle control as a central oncogenic program in EPS and identify actionable therapeutic vulnerabilities.

cancer biology↗

bioRxiv2026

AnnFlux: object-conditioned neural stochastic differential equations for single-cell perturbation dynamics

Single-cell perturbation profiling measures responses to genetic and chemical interventions, yet most models learn a static map, ignoring how populations move over time and how perturbations combine. AnnFlux, an object-conditioned stochastic differential equation, learns a drift field in latent cell-state space. Conditioning on the perturbing object makes the field queryable one object at a time, yielding per-object drifts comparable across genes and drugs. By learning a drift field tailored to each perturbation context, it interpolates a held-out timepoint in an epithelial-mesenchymal transition time course and predicts unseen perturbations. Beyond point estimates, AnnFlux improves distributional fidelity and predicts responses to held-out perturbation combinations. An IFN-response signature predicted by AnnFlux was associated with TLS proximity in an independent pan-cancer spatial atlas. This framework maps perturbation-driven cell-state evolution as continuous trajectories and represents unseen perturbations using prior-knowledge embeddings.

bioinformatics↗

bioRxiv2026

LifeSciBench: Evaluating Language Models on Realistic, Expert-Level Tasks in the Life Sciences

We introduce LifeSciBench, a benchmark of 750 expert-authored tasks designed to evaluate whether language models can handle realistic life science research work. The majority of existing life sciences benchmarks have a narrow scope or are purely knowledge-based, and therefore fail to capture the complexity of real-world research, which often involves ambiguities and requires the accurate execution of multiple dependent judgment calls. Additionally, almost all existing benchmarks span at best a small collection of subdomains within the life sciences; there is at present no existing life sciences benchmark with both the requisite breadth and depth required to convincingly measure proficiency in real-world professional research settings. LifeSciBench addresses this gap by spanning seven representative scientific workflows and seven life science domains, with each constituent task paired with a human expert-written rubric. Across five frontier and domain-specialized models, GPT-Rosalind performs best, with a task-weighted mean normalized rubric score of 0.576 and a task-weighted response pass rate of 36.1% (response-level values are first averaged within each task, and the resulting task-level values are then averaged with equal weight). LifeSciBench remains unsaturated, with 171 tasks (22.8%) having no observed passing response from any evaluated model and 261 tasks (34.8%) having a best-model pass rate below 20%. LifeSciBench therefore serves as a high-resolution evaluation of practical scientific reasoning and operational decision-making in the life sciences.

bioinformatics↗

bioRxiv2026

Human Gut Bacteria Convert Endogenous Steroids into Host Cortisol Shuttle Inhibitors

11{beta}-hydroxysteroid dehydrogenase 2 (HSD11B2) protects the mineralocorticoid receptor from glucocorticoid overstimulation by inactivating cortisol. HSD11B2 inhibition can drive receptor overactivation and contribute to hypertension; however, endogenous inhibitors remain poorly defined. Glycyrrhetinic acid-like factors (GALFs) are steroid-like metabolites that inhibit HSD11B2. Given the capacity of gut bacteria to metabolize host steroids, we hypothesized that the gut microbiome could generate GALF-like inhibitors. Here, we identify two 11-oxygenated steroid metabolites that potently inhibit human HSD11B2 in colonic cells and organoids, enabling cortisol-dependent mineralocorticoid receptor activation. We identify gut bacteria and enzymes that produce these compounds and show that their levels are markedly reduced in antibiotic-treated humans. One metabolite is elevated during pregnancy, and both are higher in individuals with stage 2 hypertension-range blood pressure. These findings reveal a bacterial route to altered host cortisol signaling and suggest a potential link between microbiome-derived GALFs and blood pressure regulation.

microbiology↗

bioRxiv2026

Extracting deep learning based morphology segmentation footprint for boar sperm cells

Reliable delineation of the head and tail of swine spermatozoa supports automated assessment of boar semen quality, from morphometric measurement to the quality control of insemination doses. In practice this relies on fluorescent staining, which adds chemistry, cost, and delay to every acquisition and labels only the nucleus. Recent work coupling imaging flow cytometry with machine learning has advanced rapidly, yet the segmentation stage still depends on a stained channel at inference and resolves the head alone. We present a supervised encoder decoder network that segments boar spermatozoa from brightfield images acquired on an Amnis ImageStream Mark II with no stain at inference. Training labels derive from the Hoechst 33342 nuclear channel (Ch7), recorded in registration with brightfield (Ch1); the dye serves only as an annotation source, and the network sees Ch1 alone. The best semantic segmentation model reaches a Dice coefficient of 0.940 on held-out cells. For comparison we evaluate a classical morphological pipeline, four further semantic segmentation models spanning three decoder families and two ImageNet-pretrained backbones, and two zero-shot pipelines built on the Segment Anything Model 2 (SAM 2), prompted either by a dilated box around the predicted head mask or by head and tail boxes emitted by a Gemma 4 Vision Language Model (VLM). The zero-shot route scores 0.637 against Ch7 but labels the tail, which the fluorescence protocol cannot. Cells scoring worst under the supervised model proved to be mostly registration failures rather than segmentation failures, as Ch7 is displaced relative to Ch1. Manual screening for this drift is infeasible at dataset scale, so we propose a flagging system that marks any Dice below 0.792, two standard deviations below the mean, and pairs it with a zero-shot pipeline in which a VLM l and SAM 2 cross-check the flagged cell before human review.

bioinformatics↗

bioRxiv2026

Structural proteomics reveals a coagulation-complement accessibility signature of macrovascular invasion in hepatocellular carcinoma

Macrovascular invasion (MVI) and extrahepatic spread (EHS) define the most aggressive, treatment-refractory hepatocellular carcinoma (HCC), yet blood-based markers that report the underlying protein-network biology are lacking. Conventional proteomics measures protein abundance but not the conformational and protein-protein-interaction (PPI) states that govern function. We applied covalent proteome painting (CPP)--a dimethylation-based accessibility assay that reads out binding-site openness--to matched tumor and serum, reasoning that intravascular tumor dissemination remodels plasma protein complexes in a manner detectable as changes in accessibility. Eight treatment-native HCC patients were profiled by CPP using matched FFPE tumor and top-14- depleted serum on a Q Exactive Orbitrap HF. The 85 tumor-serum common proteins defined an 81-protein targeted panel, validated by multiple-reaction-monitoring (MRM) mass spectrometry with heavy stable-isotope-standard peptides (296 peptides; 3,717 light/heavy transition pairs) in 22 FFPE tumors and 22 matched sera. Accessibility was the light/heavy ratio (high, open; low, closed). We assessed differential accessibility, serum-tissue translatability, pathway enrichment, and biomarker/survival performance. Aggressive disease showed broadly decreased protein accessibility. MVI-associated changes were directionally concordant between tumor and serum (Spearman {rho}=0.21; 59% concordant), driven by coagulation and complement proteins (FGG, CTSD, LBP, C4BPA); the EHS axis did not translate. Decreased-accessibility proteins were enriched for complement-coagulation cascades and IGF/IGFBP transport. A six-protein serum accessibility signature discriminated MVI (leave-one-out cross-validated AUC 0.80; best single markers ceruloplasmin 0.83 and haemoglobin- 0.77), and MVI status trended with shorter overall survival (log-rank p=0.06). Accessibility-based serum proteomics captures MVI-associated protein-complex remodeling that abundance assays miss, nominating a coagulation/complement-anchored serum signature for vascular-invasive HCC that warrants prospective validation.

systems biology↗

bioRxiv2026

A genome-wide CRISPR activation map of surface protein expression in human CD4 T cells

Surface proteins define T cell identity and function, but the abundance of each protein is not determined by transcription alone. Existing genome-wide CRISPR screens in primary human T cells either profile the transcriptome or isolate cells based on a single functional or protein phenotype. Here we present SCITO-Perturb-seq, a novel platform that couples combinatorial-indexed single-cell cytometry sequencing with pooled CRISPR activation (CRISPRa) to map the causal regulation of 201 surface proteins across 3.6 million human CD4 T cells. We find that 16% of activated genes significantly alter the expression of at least one surface protein. By applying semi-nonnegative matrix factorization to the perturbation effect matrix, we identified five modules corresponding to known CD4 T cell states. Notably, these modules group surface proteins by their shared response to perturbation, revealing coordinated regulation of proteins that are not co-expressed in unperturbed cells. SCITO-Perturb-seq represents the first genome-wide CRISPRa screen paired with direct, high-dimensional surface protein profiling, providing a comprehensive regulatory map of the CD4 T cell surface proteome.

genomics↗

bioRxiv2026

Time-resolved lineage recording reveals a pre-existing, heritable cell state underlying metastatic potential

Metastasis causes most cancer deaths1,2, yet no recurrent mutation specifically drives it3,4, raising the possibility that metastatic potential is a non-genetic yet heritable cell state. Classic experiments established that metastatically predisposed subclones pre-exist within a tumor and that these predispositions are inherited over many cell divisions5, but what molecular states or factors underlie this predisposition remain unknown. While previous lineage recording studies6,7 mapped how tumors disseminate, their recording sites saturate too quickly to resolve when a lineage branched, or to attribute a state to its founder. Here we show, using a DNA Typewriter lineage recorder8 with nearly 1,000 recording sites in lung cancer cells, that metastatic potential is already present before dissemination, with colonization predicted by a pre-existing glycolytic state and further spread by expression of ENO1, a glycolytic enzyme that also moonlights as a cell-surface plasminogen receptor9. Profiling the pre-transplant cells and the post-transplantation tumors for both their transcriptomes and their lineage recordings, we reconstructed time-resolved lineage trees across three orthotopically transplanted mice. These trees trace each liver metastasis to a single founder of known pre-transplant state, dating each dissemination event from the primary lung. When every clone was scored before transplant against 349 genes recurrently heritable in vitro, both that set and the glycolytic state independently shifted a clones odds of colonizing the lung. At the gene level, sixteen genes were both heritable and predictive of colonization, and ENO1 alone also predicted which established clones spread further. Hypoxia, the program most strongly associated with phylogenetic fitness within the metastases, did not predict colonization when scored before transplant, separating niche-selected traits from the inherited cell state. Metastatic potential in this system is therefore transmitted along the lineage rather than acquired after seeding. Looking forward, we anticipate that time-resolved lineage recorders will enable the separation of the heritable and acquired components of the cellular heterogeneity seen in single-cell studies of tumor progression and drug tolerance.

cancer biology↗

bioRxiv2026

A targeted plasma-proteomic axis separates autoimmune thyroid disease from growth-hormone deficiency

Targeted mass spectrometry (multiple-reaction monitoring, MRM) enables reproducible, multiplexed quantification of plasma proteins, but whether a fixed targeted panel can resolve endocrine disorders with overlapping systemic features is unknown. We analyzed a 256-protein targeted panel (1,894 peptides; 3,790 transitions) quantified in 57 participants spanning autoimmune thyroid disease (Hashimotos thyroiditis, n=7; Graves disease, n=5) and growth-hormone deficiency (GHD; partial, n=26; complete, n=19). Protein abundances were obtained by transition summation, log2 transformation, and per-sample median normalization. We applied unsupervised analysis (PCA, PERMANOVA), differential expression (limma), an ordered severity-trend test, and leave-one-out cross-validated classification with feature selection performed strictly inside each fold. All 256 proteins were quantified in every sample (median inter-sample r=0.875). PC1 (36% variance) separated autoimmune thyroid disease from GHD (p=0.016), whereas the global four-group structure was not significant (PERMANOVA p=0.17). No protein reached FDR<0.05, but the autoimmune-versus-GHD contrast was strongly enriched for low p-values (26 proteins at p<0.05; binomial p=5.4x10-4). The signal was biologically coherent: immunoglobulin/B-cell-receptor proteins, including CD79A, were lower, whereas proteasome subunits (PSMC5, PSMC3) and the NF-{kappa}B subunit RELA were higher in autoimmune disease. A cross-validated classifier separated the two classes (AUC 0.72; permutation p=0.05; eight proteins selected in all folds), whereas GHD severity was not predictable (AUC 0.31). A fixed 256-protein targeted panel reproducibly captures an immunoglobulin/B-cell-receptor and proteasome/NF-{kappa}B axis that distinguishes autoimmune thyroid disease from GHD but cannot resolve within-class severity.

systems biology↗

bioRxiv2026

Continuous, high-speed 3D topographic profiling of living tissue monolayers

Tissue monolayers are active living matter whose three-dimensional (3D) architecture and mechanical remodeling regulate tissue function, yet fast and continuous topographic monitoring remains difficult. Here, we present FLuorescence EXclusion microscopy fOr Monolayers (FLEXOM), a microfluidic platform that combines multilayer micropillar arrays with negative-staining optics to convert wide-field images into self-calibrated 3D height maps. Geometry-anchored reference cavities maintain in-frame calibration standards even in fully confluent monolayer fields, enabling continuous 3D profiling from isolated cells to confluent sheets with sub-second temporal resolution ([~]250 ms), sub-micrometer axial precision ([~]0.37 {micro}m), and multi-day biocompatibility and operational stability (>96 h). FLEXOM reveals two distinct anisotropic signatures in tissue monolayers. First, monolayers of different cell types show distinct vertical thickness profiles despite comparable lateral footprints. Only two vertical descriptors, mean height and height variability, distinguished cell types more accurately than nine 2D lateral descriptors, including area, circularity, and convexity (95% vs. 76%). Second, under acute osmotic shock, confluent monolayers exhibit geometrically anisotropic and temporally decoupled vertical-lateral remodeling: height changes dominate the overall volume response and precede lateral remodeling. Surprisingly, cyclic isotonic-hypotonic pulses every 3 min confine remodeling entirely to the vertical axis. Overall, our work provides a high-resolution 3D profiling tool and a framework for axis-resolved analysis of topographic responses in active living matter.

biophysics↗

bioRxiv2026

Sex and race shape serum protein biomarkers and bias diagnostic reference intervals

Clinical protein biomarkers--whether measured by FDA-cleared in-vitro diagnostic (IVD) assays or deployed as laboratory-developed tests (LDTs)--are routinely interpreted against fixed, population-agnostic reference intervals. Whether the baseline serum concentrations of these proteins differ systematically by sex and by self-reported race in healthy individuals has not been tested across a broad clinical panel on a single standardized platform. Using scheduled multiple reaction monitoring - mass spectrometry (MRM-MS) with stable-isotope-labelled internal standards, we quantified 87 proteins (39 FDA-cleared analytes; 48 LDT/research markers) in 60 healthy, age-matched donors balanced across three races (White, Hispanic, Black) and both sexes. Sex was a pervasive axis of variation: donors separated by sex in unsupervised multivariate space (PERMANOVA p = 0.020) but not by race (p = 0.14). Three FDA-target proteins differed by sex at FDR q < 0.05--sex hormone-binding globulin, haemoglobin- and albumin--with directions preserved within every race. Race differences were real but protein-specific, exemplified by ancestry-associated CD14. Critically, applying a single pooled 95% reference interval mis-flagged as many as [~]25% of a demographic subgroup as abnormal, versus the 5% expected. Demographic context should therefore be incorporated into reference intervals and decision thresholds for protein-based diagnostics.

molecular biology↗

bioRxiv2026

AI-Driven Design of Nanobinders Targeting the TSLPR Heterodimer Interface to Suppress Type 2 Inflammatory Signaling

Aberrant thymic stromal lymphopoietin (TSLP) signaling is a central driver of type 2 inflammatory diseases, yet the only approved TSLP-targeted therapy is a 150 kDa monoclonal antibody whose bulky format limits tissue penetration and precludes inhaled delivery. Here, we report an AI-driven framework for designing ultra-compact de novo nanobinders that suppress TSLP signaling by sterically disrupting assembly of the TSLPR-IL-7R heterodimer. We compare two structure-guided strategies, namely purely de novo helical bundle generation and interface-mimetic grafting of native binding motifs onto designed scaffolds. Although both yield nanomolar binders, only orthosteric mimicry of the native cytokine geometry blocks receptor heterodimerization, showing that functional antagonism is governed by precise epitope geometry rather than affinity alone. After library-based maturation, the lead nanobinder TRB5.1 is a hyper-stable monomer (Tm [~]97.4 {degrees}C) with single-digit nanomolar affinity (KD = 9.7 nM) and strict selectivity over related -chain interleukin receptors. TRB5.1 suppresses TSLP-induced JAK1 and STAT5 phosphorylation across multiple cellular models and drives a transcriptome-wide reversal of the pathogenic type 2 program. This work delivers a developable, potentially inhalable non-antibody lead and a scalable blueprint for antagonizing heterodimeric cytokine receptors.

biochemistry↗

bioRxiv2026

A targeted mass-spectrometry plasma proteome for simultaneous multi-disease screening

Population health screening is fragmented across dozens of single-condition assays. We asked whether one targeted mass-spectrometry measurement of the plasma proteome could screen simultaneously for many common diseases. Using multiple-reaction-monitoring LC-MS/MS, we quantified 126 proteins (509 peptides) directly in crude plasma from 490 individuals spanning 16 diseases (eight cancers, three inflammatory or renal conditions, and five metabolic and lifestyle disorders) plus age- and sex-matched controls, each protein referenced to a stable-isotope-labelled internal standard. Proteome variation was organised primarily by disease rather than demographics. Cross-validated classifiers detected each disease against matched controls with a mean AUROC of 0.979, a 17-way model identified the specific disease at roughly ten times chance, and a 20-protein subset reproduced full-panel performance. We show that these within-cohort estimates are optimistic upper bounds because each disease was acquired as a separate case-control batch without independent validation. We quantify this, establishing analytical feasibility and defining the validation required for translation.

molecular biology↗

bioRxiv2026

Age-dependent Structural Reorganization of the Human Plasma Proteome

Aging is characterized by progressive physiological decline, yet the molecular mechanisms underlying these changes remain incompletely understood. Although extensive work has examined age-related alterations in protein abundance, the age-dependent structural reorganization of proteins and their interaction networks remains largely unexplored. Here, we quantitatively characterize the structural dynamics of plasma proteins across healthy adult aging and identify age-dependent accessibility patterns that may serve as candidate structural biomarkers. We employed Covalent Protein Painting (CPP) coupled with multiple reaction monitoring (MRM)-based LC-MS/MS to analyze plasma from 24 healthy individuals spanning four age groups (20s to 50s, n = 6 per group). Lysine surface accessibility changes were analyzed using cosine similarity trajectory matching and limma-based linear modeling. The intersection of the two approaches defined five reproducible age-dependent trajectories encompassing 27 proteins and 57 peptides. Integrating STRING/Cytoscape network analysis with AlphaFold3 multimer complex prediction localized several CPP-detected peptides to within [~]5 [A] of predicted protein-protein interaction interfaces (KNG1, C3; REN, ALB; and ALB, FGB), suggesting that age-associated accessibility changes are organized at interaction interfaces rather than occurring as isolated local events. Our findings indicate that the human plasma proteome undergoes coordinated, network-level changes in structural accessibility during adult aging, providing a framework for structure-based biomarker discovery. Because CPP signal reflects both conformational accessibility and protein abundance, and because the cohort is modest and cross-sectional, the trajectories reported here are framed as hypothesis-generating and are interpreted with these constraints made explicit throughout.

molecular biology↗

bioRxiv2026

Anticipatory modulation of motor unit discharge rate before rapid isometric elbow flexion force production

Surface electromyography (EMG) studies have demonstrated anticipatory muscle activation prior to predictable voluntary actions. However, the mechanism through which this preparation is expressed, whether through motor-unit recruitment or discharge-rate modulation, remains to be elucidated. We employed surface decomposition EMG to quantify motor-unit behavior prior to self-paced rapid isometric elbow-flexion force pulses. Twelve healthy young men were instructed to generate elbow-flexion pulse force at 30%, 40%, and 50% of the maximal voluntary contraction (MVC) at a self-selected time. Motor-unit activity was decomposed from the biceps brachii and triceps brachii muscles, and the normalized active motor-unit number and mean discharge rate were analyzed prior to pulse onset. In the agonist, the pre-pulse increase mean discharge rate exhibited a higher value than the change in detected motor-unit count, particularly at the 40% and 50% MVC targets. The discharge-rate increase scaled with the target force, with a heightened response observed in high-threshold as compared to low-threshold motor units. Antagonist recordings with sufficient decomposition yield exhibited a similar discharge-rate-dominant pattern; however, these data were available from a smaller sample size. Present findings suggest that discharge-rate modulation is a primary motor-unit-level feature of anticipatory preparation for rapid isometric force production.

neuroscience↗

bioRxiv2026

BRAID: RT-PCR-calibrated conformal intervals for splicing ΔPSI

Differential splicing workflows usually report a {Delta}PSI point estimate and a statistical score, but these outputs do not directly state whether the RNA-seq estimate is close enough to an orthogonal validation measurement. We developed BRAID as a post-processing calibration step for splicing analyses. BRAID estimates RNA-seq {Delta}PSI from rMATS inclusion and skipping counts, retains the upstream caller evidence, and adds a 95% interval whose width is calibrated from empirical RNA-seq-to-RT-PCR residuals using split conformal prediction. The packaged differential-splicing calibrator uses a residual half-width of q = 0.341, estimated from 162 RT-PCR-validated skipped-exon events. We evaluated BRAID on three RT-PCR validation datasets covering TRA2 knockdown, mouse cerebellum versus liver, and a prostate epithelial-to-mesenchymal comparison. On the pooled common set of 139 cassette-exon events, BRAID reached 0.971 RT-PCR coverage, whereas MAJIQ, betAS, and rMATS-derived intervals reached 0.518, 0.734, and 0.633, respectively. BRAID also had the lowest pooled interval score, 0.720, compared with 2.040 for MAJIQ, 1.414 for betAS, and 1.625 for rMATS. Applying the same residual calibration to other caller outputs brought MAJIQ, betAS, rMATS, and SUPPA2 {Delta}PSI estimates close to nominal RT-PCR coverage, indicating that the gain came from interval calibration rather than from a caller-specific point estimate. In a TRA2 positive-negative validation panel, using q as a hard rMATS effect-size cutoff reduced recall, whereas using q as an interval half-width improved RT-PCR coverage. Applied to a public DM1 skeletal-muscle rMATS table, BRAID reduced 967 large-effect significant events to 68 high-confidence interval-supported events and retained known DM1 and muscle-splicing signals. BRAID provides a practical calibrated reliability layer for RNA-seq splicing studies where downstream follow-up depends on the precision of reported {Delta}PSI estimates.

bioinformatics↗

bioRxiv2026

Dissociable structural and molecular pathways of age-related change in sustained attention

Sustained attention, the capacity to maintain goal-directed attention over extended periods, declines with age but with substantial individual variability across cognitively unimpaired (CU) older adults. We sought to determine the contributions of microstructural deterioration of the superior longitudinal fasciculus (SLF), the principal white-matter tract coupling prefrontal and parietal nodes of the dorsal attention network, as well as markers of preclinical Alzheimers disease (AD) to individual differences in sustained attention in ageing. In 162 CU older adults (mean (SD) age = 73.0 (4.9), %F = 64.8) drawn from two Stanford ageing cohorts, we measured sustained attention using the gradual-onset Continuous Performance Task (gradCPT). Performance was indexed by a composite score (Att-Z) derived from discriminability (d') and response-time variability (RTV). We used diffusion MRI tractography to quantify SLF fractional anisotropy (FA) and mean diffusivity (MD) alongside two control tracts, the corticospinal tract (CST) and cingulum (CGC). NULISA multiplex plasma immunoassays were used to measured AD-related pathology (pTau-181, pTau-217), neuroaxonal injury (NfL), and astrocytic reactivity (GFAP; plasma subsample N = 146). Parallel mediation and commonality analyses tested whether structural and plasma biomarker pathways contribute independently to individual differences in sustained attention function. Older age was associated with lower Att-Z ({beta} = -0.315, p < 0.001). In simultaneous three-tract regression models, only SLF microstructure uniquely predicted Att-Z (FA: {beta} = +0.275, p < 0.001; MD: {beta} = -0.320, p < 0.001). SLF microstructure mediated the age-attention relationship (FA indirect {beta}unstandardised = -0.01, pboot <0.01, MD indirect {beta}unstandardised = -0.009 to -0.01, pboot <0.05). Plasma pTau-181 ({beta} = -0.212, pFDR < 0.03), pTau-217 ({beta} = -0.163, pFDR < 0.05), and NfL ({beta} = -0.156, pFDR < 0.05) were negatively associated with Att-Z and each mediated effects of age on performance (pTau-181 indirect {beta}unstandardised = -0.008, pboot < 0.01; pTau-217 indirect {beta}unstandardised= -0.006, pboot< 0.05; NfL indirect {beta}unstandardised = -0.011, pboot< 0.05), whereas GFAP showed no association with sustained attention in any model. SLF microstructure was negatively associated with age (FA {beta} = -0.207, p < 0.01; MD {beta} = +0.270, p < 0.001) but was not significantly associated with plasma biomarkers (all p > 0.10). In parallel mediation models, SLF microstructure and plasma pTau exhibited significant independent indirect effects on Att-Z (SLF FA indirect {beta}unstandardised= -0.010, pboot< 0.01; SLF MD indirect {beta}unstandardised = -0.009 to -0.010, pboot< 0.05; pTau-181 indirect {beta}unstandardised = -0.008 to -0.009, pboot< 0.01; pTau-217 indirect {beta}unstandardised= -0.006 to -0.007, pboot< 0.05). Effects of NfL were attenuated when combined in a model with pTau (indirect {beta}unstandardised = -0.005 to -0.008, pboot> 0.15), suggesting shared variance among these markers. These findings reveal multiple pathways that contribute to individual differences in sustained attention in ageing that are detectable before clinical impairment, including SLF white matter integrity -- which is linked to the structural disconnection of the dorsal attention network -- and preclinical AD pathology. The independence of these pathways identifies two separable biological targets for preserving attentional capacity in CU older adults.

neuroscience↗