Search bioRxiv⌕ Search

Biology subjects

Shi, Z.

Publications and source records attributed to Shi, Z..

At least 37 records · Page 2Linked to original sources

Context-Dependent Modulation of Duration Encoding in the Supplementary Motor Area

Duration judgments are calibrated by the statistics of recent temporal experience, but how this contextual influence is expressed across human timing-related brain regions remains unclear. We used functional MRI while 24 participants performed a visual temporal bisection task under two distributional contexts: one biased toward longer durations and one biased toward shorter durations. Behaviorally, distributional context shifted subjective duration judgments without changing in temporal precision, consistent with a prior-related bias. Neurally, duration-related BOLD responses varied across a priori timing regions as a function of context. The clearest region-specific evidence was stronger left-insula duration modulation in the short-biased context. SMA/pre-SMA contributed to the broader timing-network pattern, and its context-specific asymmetry was sensitive to the alignment of the duration regressor. Exploratory SMA-seeded connectivity analyses provided converging evidence that context also modulated duration-related coupling with timing- and attention-related regions. Together, these findings indicate that distributional context impacts the relative weighting of established timing-related regions during duration judgments, rather than recruiting anatomically separate timing systems.

neuroscience↗

Neural Substrates of Duration Serial Dependence: Opposing Modulation in Basal Ganglia and Posterior Medial Cortex

Perceptual judgments are systematically biased toward recent experience, a phenomenon called serial dependence. The brain circuits underlying this effect in time perception remain unknown. Using fMRI, we identified the neural substrates of duration serial dependence and tested whether they overlap with those for motion direction. Twenty-seven participants reproduced either the duration or direction of a visual stimulus; a retro-cue revealed which feature to report 3-s after stimulus onset, separating memory-trace signals from task-switching signals. Duration serial dependence was robust and roughly four times stronger on task-repeat than task-switch trials; direction serial dependence was near zero. The previous duration produced opposing responses in the bilateral putamen (suppression) and in the posterior cingulate cortex and precuneus (reinstatement); no analogous signal emerged for direction. Task switching from direction to time task engaged the inferior frontal junction and the dorsolateral prefrontal cortex, which carried no serial dependence signal. The anatomical dissociation resolves the behavioral pattern: the duration trace is encoded in the putamen and posterior medial cortex on every trial, but task-set reconfiguration by the frontoparietal network gates its read-out into behavior, explaining the task repetition effect. These results reveal a dissociation between two functionally distinct circuits: the putamen and posterior cingulate/precuneus carry signals from the previous trial, whereas frontoparietal regions support task switching but carry little serial dependence information.

neuroscience↗

Environmental Volatility Shifts Visual Search from Capture to Caution

Real-world distractors occur in environments whose states change at different rates. We asked whether such volatility alters early attentional gating or instead changes the criterion for committing to a response. Observers performed an additional-singleton search task with concurrent eye tracking while distractor presence followed high- or low-volatility sequences, with overall distractor prevalence held constant. Trial-pooled oculomotor capture was higher under high volatility, a pattern that appears to indicate altered filtering. That inference did not survive repetition-aware analysis: once the same-location run position was matched, capture did not detectably differ across volatility regimes. The pooled capture effect was therefore consistent with a structural consequence of the volatility manipulation, which enriched high-volatility blocks with early-run positions where capture is intrinsically high. The positive volatility signature appeared on distractor-absent trials, where high-volatility blocks were associated with longer target latency, more fixations, longer final-target dwell, and fewer errors. Same-location repetition learning showed no detectable difference in slope across regimes. A hierarchical drift-diffusion model (DDM) and a complementary volatility Kalman-filter (VKF) dynamic-state comparison indicated that manual responses were better described by architectures that allow both boundary-related and drift-related components than by a boundary-only account. Volatility, therefore, did not show detectable evidence of impairing the local gating rule; instead, the converging evidence points to a post-selective verification/caution profile, consistent with a precision-weighted read-out of environmental uncertainty.

neuroscience↗

DMN Connectivity Predicts Location-Learning-Related Attentional Capture Vulnerability

Learned spatial regularities can reduce attentional capture by salient distractors, but observers vary in how strongly such regularities shape distractor interference. We asked whether this variability is associated with default mode network (DMN) functional connectivity, a network measure linked to individual differences in attention, internally guided cognition, and contextual memory. Using an existing fMRI visual-search dataset (N = 33), we tested whether task-run DMN connectivity was related to capture by distractors assigned to a learned spatial-probability history. Participants belonged to two training groups that differed in which distractor dimension carried the spatial bias: same-dimension distractors in one group and different-dimension distractors in the other. This design allowed us to ask whether the DMN association followed the design-defined biased dimension rather than general distractibility or a fixed physical distractor feature, while considering that the biased dimension also involved greater exposure. Higher within-DMN connectivity was associated with larger capture scores for the biased distractor dimension, whereas capture by the unbiased dimension showed little corresponding relation. This association remained positive after controls for inter-trial priming, head motion, nuisance signals, task-regressor residualization, global-signal regression, motion scrubbing, and trial-level exposure history. Exploratory edge-level analyses provided descriptive convergence, showing that capture-associated positive edges were concentrated around the DMN. These findings suggest that DMN connectivity may relate to individual differences in vulnerability to distractors embedded in learned spatial-probability histories.

neuroscience↗

Pre-stimulus Cortical State Modulates Dimension-Specific Attentional Capture

Suppressing visual distraction depends on observers goals and learning, but also on the brain state in which the next display arrives. We reanalyzed an fMRI visual-search dataset (N = 34) to ask whether pre-stimulus global signal (GS), a broad state-sensitive BOLD measure, predicts performance and attention-network engagement during learned distractor suppression. Participants searched for an orientation target while ignoring salient distractors defined in the same orientation dimension or a different color dimension; across groups, spatial regularities were assigned to same- or different-dimension distractors. Behavioral and fMRI contrasts reproduced the expected functional architecture: same-dimension distractors slowed responses and recruited a frontoparietal capture-related network, whereas different-dimension distractors produced little interference. Learned spatial suppression was clearest for same-dimension distractors, which produced larger costs and stronger visual and frontoparietal responses in the rare than frequent region. Within this architecture, GS predicted performance in opposite directions across learning contexts: higher GS preceded slower responses when same-dimension distractors carried the spatial regularity, but faster responses when different-dimension distractors did. Higher GS also preceded weaker same-dimension distractor-evoked frontoparietal activation, without a reliable group difference in that neural effect. Default-mode-network activity showed the same behavioral pattern as GS, whereas pupil size showed weaker group-level convergence and its RT slopes did not correlate with GS-RT slopes across participants. Learned distractor suppression therefore appears embedded in a broader state-performance mapping, linking pre-stimulus cortical state to how learned priority settings guide the next trial.

neuroscience↗

Differential effects of α-Synuclein monomers and seeds on the material properties of Tau condensates

Tau and -Synuclein (Syn) frequently co-aggregate in various neurodegenerative disorders. Recently, Tau has been shown to form dynamic, liquid-like condensates that can recruit Syn, and potentially serve as a precursor to pathological aggregation. However, the quantitative impact of Syn on the material properties of these condensates remains elusive. Here, we measure the viscosity and interfacial tension of Tau condensates and determine how these properties are modulated by Syn monomers and fibril seeds. We find that while both forms of Syn partition efficiently into Tau condensates, they exert vastly different effects on the condensates material state. The viscosity of Tau condensates remains unchanged in the presence of Syn monomers at concentrations up to 200 {micro}M, accompanied by a moderate reduction in the condensates interfacial tension. In contrast, the addition of only 5 {micro}M Syn fibril seeds triggers rapid solidification of Tau condensates, manifested by a nearly 100-fold increase in condensate viscosity within one hour. These findings provide quantitative insights into condensate mechanics, highlighting the unique capacity of Syn seeds to drive the liquid-to-solid transition of Tau condensates that may underlie the formation of pathological aggregates.

biophysics↗

Biomolecular condensates provide a unique environment for redox-mediated protein crosslinking

Biomolecular condensates, often formed through liquid-liquid phase separation, are dynamic cellular compartments. Here, we demonstrate that a wide range of fluorescently tagged proteins undergo inadvertent, condensate-mediated crosslinking, resulting in rapid solidification of condensates under common fluorescence imaging conditions. The process is driven by excitation-induced, short-lived reactive oxygen species (ROS), whose otherwise limited crosslinking potential becomes uniquely enabled in the dense phase. In live cells, excitation-induced ROS potently trigger stress granule formation, while the ROS-driven solidification of condensates is modulated by compartment-dependent antioxidant buffering. Our findings demonstrate that condensates create a distinct environment that enables ROS chemistry unlikely to occur in the bulk cytosol. Furthermore, the cellular redox level can be a general regulator of condensate rheology. Beyond biological insights, our findings underscore the need for scrutiny when examining fluorophore-labeled condensates.

biophysics↗

Lipoengineering of Biomolecular Condensates Controls Material Properties and Multiphase Hierarchy to Guide Organoid Morphogenesis

Cells use post-translational modifications (PTMs) to reconfigure biomolecular condensates across length scales, space, and time.1,2 While charged PTMs are well-known electrostatic switches,3,4 how ubiquitous neutral PTMs shape condensate plasticity and hierarchy remains unclear. Here, we establish a set of design principles for using site-specific lipidation, a class of neutral hydrophobic PTMs, to rationally control properties and interactions of engineered biomolecular condensates. Through systematic analysis of over 80 lipidated synthetic intrinsically disordered proteins (IDPs), we uncovered two distinct axes of control. First, the interplay between the lipid and the local three-residue sequence of its attachment site acts as a programmable switch for cohesion--the homotypic interactions that define the material state of the condensed phase-- directing assemblies toward dynamic liquids, arrested gels, or ordered fibrillar solids. Second, the lipid, together with the global properties of the IDP scaffold, tunes adhesion--the heterotypic interactions that govern condensate miscibility and hierarchical organization. We harnessed these principles to rationally engineer complex, multi-phase architectures and create hybrid hydrogels with programmed microstructure and material properties that guide the morphogenesis of functional intestinal organoids. These findings establish a new framework for lipoengineering advanced biomaterials and provide a blueprint for dissecting structure-property relationships across diverse classes of PTMs.

bioengineering↗

Overlapping MHC class I/II Epitopes Program cDC1-like Differentiation of Monocyte-Derived Dendritic Cells via mTORC1 Signaling Inhibition

Viral infection polarizes monocyte-derived dendritic cells (moDC) to initiate type 1 immunity. The availability of overlapping (homologous) MHC class I and II epitopes, an occurrence frequently and primarily associated with intracellular infection, significantly enhances this process; however, the underlying mechanism(s) are unclear. We demonstrate that moDC loaded with homologous MHC epitopes acquire a cDC1-like phenotype in a process governed by mTORC1. mTORC1 pathway inhibition leads to NF-{kappa}B-mediated expression of IL-12 and other type I immune polarizing genes. The observed cDC1-like gene signature was also significantly enhanced in clinical moDC vaccine products made through methodologies that enforced class I and II antigenic homology. Collectively, these findings reveal a novel and previously unrecognized mechanism of immune governance that might also be exploited in cancer immunotherapy. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=167 SRC="FIGDIR/small/716309v1_ufig1.gif" ALT="Figure 1"> View larger version (41K): org.highwire.dtl.DTLVardef@15475borg.highwire.dtl.DTLVardef@ffe5fborg.highwire.dtl.DTLVardef@53c761org.highwire.dtl.DTLVardef@46c4ef_HPS_FORMAT_FIGEXP M_FIG C_FIG

immunology↗

PRISM: A High-Throughput Simulation Infrastructure for CADD Agents

Despite rapid progress in AI agents for computer-aided drug design (CADD), protein-ligand simulation workflows remain fragmented across disparate tools, creating a major bottleneck for scalable candidate evaluation. Here, we present PRISM (Protein-Receptor Interaction Simulation Modeler), a Python platform built on GROMACS that unifies ligand parameterization across multiple force fields, automated system construction, enhanced sampling, multi-tier binding free energy estimation, and trajectory analysis within a single workflow. Through the Model Context Protocol (MCP), PRISM further serves as the computational infrastructure for CADD-Agent, an expert-workflow-driven AI agent designed to orchestrate hierarchical drug screening pipelines. As a pilot application, we applied PRISM to riboflavin synthase and demonstrated end-to-end automation from candidate library assembly to binding pocket characterization, identifying a potential allosteric inhibition site at the oligomerization interface. Together, these results establish PRISM as a high-throughput simulation infrastructure for agent-enabled CADD.

biophysics↗

Spatio-temporal mapping of immune cell dynamics during human sequential lymph node metastasis

Regional lymph node (LN) metastasis critically influences distant metastatic progression, anti-tumour immunity, and patient prognosis. While tumour-induced immune modulation in tumour-draining LNs (TDLNs) has been extensively studied using murine models, the systematic reconstruction of the immune system from primary tumours through TDLNs and subsequent lymph nodes in human cancer progression remains understudied. Here, we utilised integrated multi-omics approaches, including imaging mass cytometry, single-cell RNA sequencing, Visium and Xenium spatial transcriptomics, and multi-colour immunofluorescence to systematically characterise immune cell dynamics across 147 paired primary tumours, sentinel TDLNs (S-TDLNs), and secondary axillary LNs (ALNs) obtained from 50 treatment-naive triple-negative breast cancer patients with different progression statuses. Our comprehensive profiling revealed critical immune alterations, such as decreased type-2 conventional dendritic cells (cDC2), naive T cells, and B cells, along with an increase in immunosuppressive macrophages. Developing a novel single-cell transformer model, we identified substantial alterations in various immune cell populations, notably MARCO+ macrophages, which strongly correlated with breast cancer patient survival outcomes. Spatial analysis combined with our newly integrated cell-cell interaction platform revealed diminished immune cell communication and impaired priming interactions among dendritic cells, B cells, and T cells within metastatic lymph nodes and primary tumour sites. In an independent neoadjuvant immunotherapy cohort of 36 TNBC patients with 52 lymph node samples, we found preservation of CD1c cDC2 in lymph nodes predicted pathological complete response and longer event-free survival, highlighting cDC2 as a potential biomarker and therapeutic target. Collectively, this systematic mapping of immune landscape alterations during human sequential LN metastasis provides essential insights for understanding cancer metastasis mechanisms and paves the way for innovative immunotherapeutic strategies.

immunology↗

Machine Learning Ensemble Reveals Distinct Molecular Pathways of Retinal Damage in Spaceflown Mice

BackgroundSpaceflight-associated neuro-ocular syndrome (SANS) poses significant risks to astronaut visual health during long-duration missions, yet its underlying molecular mechanisms remain incompletely understood. Oxidative stress and apoptosis are candidate molecular drivers, but their transcriptomic signatures in spaceflight-exposed retinal tissue have not been systematically characterized. MethodsWe applied a machine learning ensemble of linear regression models to predict two ocular phenotypes: 4-hydroxynonenal (4-HNE) immunostaining as a marker of lipid peroxidation-mediated oxidative damage; and TUNEL positivity as a marker of apoptotic cell death. In this observational study, we use bulk retinal gene expression data obtained from a controlled experiment with ground control and spaceflown mice to predict these phenotypes. Gene Ontology pathway enrichment was performed on the most predictive gene sets for each phenotype. ResultsThe 4-HNE phenotype was predicted by genes that converge on membrane-associated pathways, photoreceptor protein modification, synaptic dysfunction, and extracellular matrix dysregulation, including B2m, Tf, Cnga1, mt-Nd1, Snap25, and Efemp1. The genes predicting the TUNEL phenotype revealed a distinct signature emphasizing stress-induced apoptosis, rod photoreceptor degeneration, and endoplasmic reticulum dysfunction, with Ddit4, Nrl, Rom1, Reep6, and Gabarapl1 emerging as central regulators. ConclusionsOxidative lipid peroxidation and apoptotic cell death represent complementary and molecularly distinct pathological mechanisms in spaceflight-exposed murine retinal tissue. The gene signatures provide a putative molecular framework for developing noninvasive biomarkers and therapeutic targets to monitor and protect astronaut visual health during long-duration and deep-space missions.

bioinformatics↗

Functional In-Context Learning in Genomic Language Models with Nucleotide-Level Supervision and Genome Compression

AO_SCPLOWBSTRACTC_SCPLOWGenomic foundation models aim to learn general-purpose representations directly from DNA sequence, enabling sequence understanding, generation, and probabilistic reasoning across a wide range of biological tasks. Scaling such models to genomic lengths, however, remains challenging due to the tension between long-range context, nucleotide-level resolution, and practical computational efficiency. Architectural innovations have enabled increasingly long nominal inputs, but often struggle to translate additional context into meaningful performance gains, particularly in the presence of sparse functional signal along eukaryotic genomes. In this work, we revisit the design of long-context genomic foundation models from the perspective of training objective and data construction. We introduce Factorized Nucleotide Supervision (FNS), which reconciles efficient k-mer tokenization with single-nucleotide likelihoods through probability marginalization, and Genome Compression Pretraining (GCP), which reshapes the training distribution by concentrating on gene-centric and regulatory regions. Together, these techniques enable standard transformer-based models to perform functional in-context learning without sacrificing nucleotide-level fidelity or computational efficiency. Building on these ideas, we present a family of autoregressive genomic foundation models supporting contexts of up to 98k base pairs across eukaryotic and prokaryotic genomes. Across training-free evaluations and downstream fine-tuning benchmarks, our models consistently improve over prior approaches and match or exceed state-of-the-art baselines while enabling substantially more efficient inference. Together, these results demonstrate that aligning supervision and data regimes with the biological structure of genomic sequence provides a principled and effective path toward scalable and biologically faithful genomic language modeling. Models, data, and scripts for downstream analyses are publicly available at https://huggingface.co/GenerTeam.

genomics↗

Discovery of potent oligopeptides for various metabolic diseases using deep learning

Artificial intelligence (AI)-based methods are increasingly critical in peptide drug discovery, but their applications are limited in a narrow scope such as antimicrobial peptides. For indications where peptide therapy inherently excels, such as metabolism, endocrinology, and tissue regeneration, AI-based pipeline for therapeutic peptide discovery is long awaited but still unmet. Here, we propose a deep learning-based pipeline, Deepeptide, capable of identifying therapeutic oligopeptides for various metabolism-related indications. Leveraging the intrinsical relationship of disease indication -- biological processes -- molecular functions, Deepeptide discovers oligopeptides with indication-ameliorating-related molecular functions as lead candidates for indication of interest. Deepeptide was applied in five representative indications of metabolism, endocrinology, and tissue regeneration: angiogenesis, lipid metabolism, osteogenesis, glucose metabolism, and anti-angiogenesis. Overall, 62% of the identified oligopeptide candidates demonstrated significant bioactivity in vitro, with most of them showing comparable potency to the first-line drugs. Notably, the heptapeptide AP7 exhibited angiogenic potency comparable to VEGF in excisional wound splinting mouse model by promoting cell migration rather than proliferation, and hexapeptide TP6 showed significant dual-efficacy against hyperlipidemia and obesity in high-fat diet mice by inhibiting lipid synthesis and regulating gut microbiota. These findings highlight the potential and generalizability of Deepeptide in therapeutic oligopeptide discovery for metabolic diseases.

bioinformatics↗

PARP10 condensation inhibits viral infection via targeting NAD+ homeostasis

ADP-ribosylation is a critical post-translational modification mediated by diphtheria toxin-like ADP-ribosyltransferases (ARTDs). The human ARTD family comprises 17 members that catalyze either poly- or mono-ADP-ribosylation. However, the functions and regulatory mechanisms of many ARTD proteins remain poorly understood. Here, we uncover an antiviral role for PARP10 through its ability to form biomolecular condensates. Viral infection triggers PARP10 condensation, a process driven by multivalent homotypic interactions among the structured domains. We show that mono-ADP-ribosylation of PARP10 suppresses its condensation, serving as a negative feedback mechanism that regulates condensate dynamics and protein stability. PARP10 condensates exhibit enhanced enzymatic activity, leading to decreased NAD+ levels and disruption of NAD+ homeostasis, ultimately inhibiting viral replication. Our findings establish PARP10 condensation as a novel mechanism for ARTD enzyme compartmentalization, with significant implications for innate immunity and host defense.

cell biology↗

Multi-timescale Rhythmic Dynamics in Rostral Ventromedial Medulla Neurons

Effective pain therapies increasingly target neural circuits that regulate nociceptive processing; yet, how descending control systems regulate pain across time remains poorly understood. Because pain regulation must coordinate rapid defensive responses with slower fluctuations in physiological state, these neural circuits are likely to operate across multiple timescales. However, whether such dynamics exist in brainstem pain-control circuits remains largely unknown. Here, we investigated this question in populations of rostral ventromedial medullary (RVM) pain-modulating neurons. The RVM contains ON- and OFF-cells that exert descending control over spinal nociceptive transmission, regulating pain sensitivity and behaviors. By integrating neuronal recordings with probabilistic modeling, we show that unstimulated and stimulus-driven conditions give rise to distinct timescales of ON- and OFF-cell dynamics. During noxious stimulation, we find that population responses undergo rapid activation followed by superimposed slow and fast recovery dynamics over tens of seconds. In contrast, the same neurons exhibit quasi-periodic fluctuations in firing activity on the order of minutes in the absence of stimulation. Gaussian-process models show that these slow dynamics are statistically predictable from past activity, indicating structured temporal organization beyond stimulus-evoked responses. Taken together, these results indicate that descending pain-control circuits exhibit structured dynamics spanning rapid pain-related signaling and slower fluctuations associated with ongoing physiological state. SignificancePain regulation requires coordination between rapid defensive responses and slower changes in physiological state, yet how these processes are integrated in the brain remains unclear. We show that neurons in a key brainstem pain-control center, the rostral ventromedial medulla, operate across multiple timescales. Using neuronal recordings and computational modeling, we find that these neurons exhibit both fast responses to painful stimuli and slow, structured fluctuations in ongoing activity. These results demonstrate that descending pain control is temporally organized beyond immediate stimulus-evoked responses. This provides a framework for understanding how pain is regulated over time.

neuroscience↗

Dysregulation of transcription networks regulating oligodendrogenesis in age-related decline in CNS remyelination

In demyelinating diseases like multiple sclerosis (MS), efficient remyelination is critical for functional recovery. Remyelination efficiency declines with age, and is linked to progressive disability. The gene regulatory network underlying remyelination, and how it is altered with aging, remains unclear. Here we present a comparative single-nucleus RNA and ATAC sequencing analysis of remyelination in young and aged mice. We identified gene modules dynamically expressed throughout oligodendrocyte differentiation, revealing age-dependent changes in key processes related to myelination. Multi-omic analysis allowed us to map the regulatory network driving efficient remyelination within oligodendrocyte lineage cells in young mice. We highlight key transcription factors in the network dysregulated with age, and we describe similar dysregulations in MS lesions. Modifying the expression of these transcription factors in primary oligodendrocyte progenitor cells impacts differentiation. These findings provide a foundational understanding of this regenerative process in the context of aging and in chronic demyelinating diseases.

neuroscience↗

Bronsted-basic small molecules activate GTP hydrolysis in Ras Q61 mutants

The RAS oncogenes (KRAS, HRAS, NRAS) are among the most frequently mutated genes in human cancer, affecting over three million patients annually. Therapeutic development has largely focused on inhibitors for KRAS codon 12 mutations (G12C/D/V/S/R) which are key drivers in lung, colorectal, and pancreatic cancers. In contrast, mutant-selective inhibitors for Q61 variants remain elusive. A common mechanistic feature of G12 and Q61 mutants is the impaired hydrolysis of GTP, which traps Ras in its active, signaling-competent state. We envisioned that an alternative therapeutic strategy - reactivation of GTP hydrolysis - could address this shared oncogenic mechanism. Here we report small molecules that accelerate GTP hydrolysis in K-Ras Q61 mutants. These compounds compensate for the loss of the catalytic residue Gln61 by introducing a general base into the active site, selectively enhancing hydrolysis of K-Ras Q61X (X = H, L, K, R) mutants by up to 20-fold while sparing the wild-type protein. In mutant cancer cell lines, these compounds reduce GTP-bound Ras levels and suppress downstream signaling. We show that the chemical design principles are generalizable to other Ras isoforms. This work establishes a mechanistic foundation for small-molecule "GTPase activators" and offers a new paradigm for targeting Ras-driven cancers.

cancer biology↗