Biology subjects

Zang, Y.

Publications and source records attributed to Zang, Y..

At least 19 recordsLinked to original sources

bioRxiv2026

BioMod: Biological Module Tokenization for Single-Cell Foundation Modeling

Single-cell foundation models typically represent cells as flat sequences of individual genes, requiring higher-order biological organization to be inferred implicitly from sparse expression profiles. Similar to the evolution from character-level to semantic tokenization in natural language processing,we hypothesize that cellular states can be more efficiently represented using biologically meaningful intermediate units. Here we present BioMod, a 50-million-parameter single-cell foundation model that introduces biological modules as representation units between genes and cells. BioMod constructs an overlapping vocabulary of 660 biological modules derived from curated pathways and data-driven gene programs, replacing thousands of gene-level inputs with compact module representations. Pretrained on approximately 73 million human cells using high-ratio module masking with an asymmetric decoder, BioMod achieves state-of-the-art performance across cell-type annotation, zero-shot batch integration, perturbation-response prediction and drug-response forecasting while maintaining a compact architecture that enables efficient inference on a single GPU. Module-level analyses further demonstrate that BioMod captures biologically meaningful cellular programs and preserves pathway-level organization.

bioinformatics↗

bioRxiv2026

Internalized Components of Membrane Attack Complexes Disrupt Proteostasis and Acquire Alarmin-Like Properties

Immune effects of membrane attack complexes (MAC) have been widely attributed to their abilities to cause cell death. Here, we show that the MAC component, C9, forms non-cytolytic aggregates with pro-inflammatory effects. Intracellular aggregates of C9 are detected within inflamed tissues of patients in association with endothelial cell (EC) activation but not increased cell death. We identify NUMBL as a Rab35 effector that directly binds surface-bound C9 to promote C9 internalization and entry into the endolysosomal pathway. Within acidified endolysosomes, C9 forms insoluble aggregates that are targeted for degradative aggrephagy in a process that activates NF-{kappa}B. For C9 aggrephagy to occur, ZFYVE21, a Rab5 effector, complexes with RNF34 to bridge C9 aggregates to LC3B+ aggresome membranes. We detect C9 aggregates in vivo, and we show that a ZFYVE21-RNF34 signaling axis is required for C9 aggrephagy and NF-{kappa}B -dependent EC activation in three separate mouse models. Mice with conditional loss of ZFYVE21 in ECs show reduced aggregraphy, resulting in attenuated systemic inflammation and reduced tissue injury following skin transplantation. Our data show that the C9 component of MACs forms intracellular aggregates with alarmin-like properties.

biochemistry↗

bioRxiv2026

A Tripartite Map of the Ventromedial Prefrontal Cortex

The ventromedial prefrontal cortex (VMPFC) has been repeatedly implicated in affect, valuation, and social cognition, yet how these diverse functions are organized within a single cortical territory has remained unresolved. Here, we integrate large-scale meta-analysis, individual-level task fMRI, artificial neural-network encoding models, and multimodal connectivity analyses to reveal the internal functional architecture of the human VMPFC. Across four complementary studies, we identify a robust tripartite organization along the anterior-posterior axis, comprising posterior affective, middle valuation, and anterior social functional motifs. Connectivity fingerprinting demonstrates that each motif is preferentially embedded within distinct large-scale brain networks, providing a mechanistic account of VMPFC functional specialization. This organization is reproducible at the level of individual subjects, generalizes to naturalistic stimuli, extends across development, and shows cross-species correspondence with non-human primates and multiple neurobiological markers. Together, these findings resolve a long-standing organizational question and establish a biologically grounded framework for interpreting VMPFC function.

neuroscience↗

bioRxiv2026

IL-36γ Drives Th9 cell Differentiation via IκBζ to Sustain Antitumor Immunity

Th9/Tc9 cells polarized with TGF-{beta} and IL-4 exhibit superior antitumor efficacy, prompting extensive efforts to optimize their differentiation protocols and augment therapeutic potency. In the current study, we identified IL-36{gamma} as a potent cytokine that synergizes with TGF-{beta} to robustly drive Th9 differentiation both in the presence and absence of IL-4. IL-36{gamma}-programmed Th9 cell subsets exhibited phenotypic and transcriptional profiles identical to classic Th9 cells. Mechanistically, IL-36{gamma} drove Th9 cell differentiation through amplifying key signaling pathways such as STAT6, STAT5, and NF-{kappa}B that are essential for classic Th9 programming. Notably, we uncovered a novel regulatory axis wherein IL-36{gamma} upregulates the transactivation factor I{kappa}B{zeta}, which directly governs Th9 lineage specification. In in adoptive cell therapy (ACT) models, IL-36{gamma}-polarized Th9 cell subsets demonstrated enhanced antitumor efficacy, attributable to their sustained persistence, reduced exhaustion markers and stem-like/memory properties. Collectively, this study elucidates a previously unrecognized I{kappa}B{zeta}-dependent mechanism underpinning Th9 differentiation and highlights the translational potential of IL-36{gamma}-engineered Th9 cells as a valuable ACT strategy for refractory tumors.

immunology↗

bioRxiv2026

Cerebellar Microcircuits and Cortico-Cerebellar Cooperation Enable Robust Decisions

The cerebellum is increasingly implicated in perceptual decision-making, yet how cerebellar circuits-- without prominent recurrent excitation and slow reverberant loops--could support sparse evidence accumulation over behavioral timescales remains unclear. We present a biologically constrained modeling framework showing that cerebellar microcircuits can implement graded accumulation and competition without cortical-like excitatory recurrence. Type-II Purkinje-neuron excitability generates firing-rate hysteresis that prolongs the impact of brief inputs far beyond intrinsic membrane and synaptic time constants, enabling accumulation across long inter-event intervals. Purkinje neuron collateral inhibition produces competitive divergence and tunes temporal evidence weighting, revealing a trade-off between commitment and primacy bias. In a bidirectionally coupled cortico-cerebello-cortical model, cerebellar processing reduces primacy while cortical processing reduces indecision, improving robustness. Together, these results propose a mechanistic division of labor that positions the cerebellum as an active computational partner in perceptual decisions.

neuroscience↗

bioRxiv2026

Balancing Inhibition and Sparsity for Stable, Accurate Cerebellar Learning

The cerebellums structured circuitry supports learning across motor and cognitive domains, yet the coding strategies in granule cells that enable this versatility remain unclear. Using a theoretical cerebellar model, we examined how feedforward inhibition (FFI) and feedback inhibition (FBI) shape granule cell activation patterns to optimize learning in two tasks: complex trace learning and pattern identification. For trace learning, performance depends on spatiotemporally ordered granule cell activity shaped by FBI, with temporal sparsity emerging as the key determinant of accuracy. For pattern identification, both pathways support high accuracy, with only slight sensitivity to pathway choice. In both tasks, spatial sparsity becomes critical in incremental learning to prevent memory interference, a role reinforced by advanced synaptic plasticity strategies. These findings identify sparse granule cell activation as a unifying principle for cerebellar learning and reveal task-dependent roles of inhibitory pathways, providing a mechanistic framework for understanding stability-plasticity trade-offs in cerebellar learning.

neuroscience↗

bioRxiv2026

Social Functioning in Autism: A Systematic Review and Meta-analysis

Atypical social functioning is a core feature of autism, yet findings remain fragmented across components and development. We aimed to systematically integrate this literature and characterize the organization, development, and moderators of social functioning in autism. We conducted a systematic review and meta-analysis of behavioral studies published between January 1990 and August 2025, identified through PubMed, Web of Science, and prior reviews, including studies with clinically diagnosed autistic individuals and neurotypical controls. A qualitative synthesis and two complementary quantitative meta-analyses were performed, with risk of bias evaluated through study-level characteristics. A total of 2,622 studies (94,114 autistic and 172,847 neurotypical individuals across 32 countries) were included, covering 22 social components that clustered into five domains. Overall group differences were substantial (Hedges g = -0.744, 95% CI [-0.797, -0.690]). Differences emerged earliest in motivation-based processes ([~]6 months), followed by motor, emotion, and inference domains, and showed age-related divergence alongside improvement in some skills. Cross-domain analyses revealed stronger interdependencies in autism and an organizational pattern most consistent with serial relationships among domains. These findings should be interpreted in light of methodological heterogeneity, underpowered samples, and uneven cultural representation. Together, the results provide an integrative framework for understanding the organization and development of social functioning in autism, with implications for precision subtyping, developmentally timed interventions, and neurodiversity-informed research and policy. This study was pre-registered (PROSPERO: CRD42024566141).

developmental biology↗

bioRxiv2026

Cerebellum-Inspired Kernel for Robust OOD Detection

Detecting novel stimuli is a fundamental neural function, yet its machine learning counterpart--out-of-distribution (OOD) detection--remains challenging, with models often making overconfident predictions on unseen inputs. Inspired by the strong pattern-separation capabilities of cerebellum-like circuits, we introduce a cerebellum-inspired kernel with an efficient closed-form implementation. Combining random Gaussian projection with Top-k sparsification, the kernel reshapes similarities in high-dimensional space to enhance separability between in-distribution (ID) and OOD samples. On OpenOOD benchmarks, our kernel consistently improves multiple baseline methods, and pairing it with the energy score achieves performance comparable to or exceeding current state-of-the-art approaches. The closed-form design also avoids the high computational cost of large-expansion explicit mapping. These results demonstrate the generality and potential of cerebellar kernels for OOD detection and other tasks requiring efficient pattern separation under limited computational resources.

neuroscience↗

bioRxiv2026

Decomposing multi-scale dynamic regulation from single-cell multiomics with scMagnify

Deciphering the highly coupled regulatory circuits that drive cellular dynamics remains a fundamental goal in biology. However, capturing the multi-scale time-lagged dynamics and combinatorial regulatory logic of gene regulation remains computationally challenging. Here we present scMagnify, a deep-learning-based framework that leverages multiomic single-cell assays of chromatin accessibility and gene expression via nonlinear Granger causality to reconstruct and decompose multi-scale gene regulatory networks (GRNs). Benchmarking on both simulated and real datasets demonstrates that scMagnify achieves superior performance. scMagnify employs tensor decomposition to systematically identify combinatorial TF modules and their activation profiles across different time-lags. It enables a hierarchical dissection of the regulatory landscape-- from the activity of individual regulator to the combinatorial logic of regulatory modules and intercellular communications. We applied scMagnify to human hematopoiesis and mouse pancreas development, where it successfully recovered known lineage-driving regulators and provided novel insights into the combinatorial logic that governs cell fate decisions. Furthermore, in the context of kidney injury, scMagnifys intracellular communication module mapped key signaling-to-transcription cascades linking microenvironment cues to pathological epithelial cell changes. In summary, scMagnify provides a powerful and versatile computational framework for dissecting the multi-scale regulatory logic that governs complex biological processes in development and disease.

bioinformatics↗

bioRxiv2026

Set-up, validation, evaluation, and cost-benefit analysis of an AI-assisted assessment of responsible research practices in a sample of life science publications

The (semi-)automated screening of publications for diverse quality and transparency criteria is at the core of systematic literature assessment. Typically, the assessment process involves two initial reviewers and one additional reviewer for cases that require reconciliation. Here, we explore to what extent this process can be assisted by Large Language Models (LLMs). Specifically, whether LLMs are capable of assessing responsible research practices (RRPs) in scientific papers in a robust way. We employed proprietary LLMs to assess an initial set of 37 papers across ten RRPs. The same papers were also reviewed by three human reviewers. We iteratively redesigned prompts to increase model accuracy compared to human ratings which we treated as the gold standard. The resulting pipeline was validated on an additional set of 15 papers. We show that LLM accuracy is comparable to single human reviewer performance (90% for LLM vs 86% for a single human reviewer). However, performance strongly depended on the specific RRPs with accuracy ranging from 40% to 100%. LLMs exhibited an affirmative bias, making more errors when practices were not reported in the papers. Overall, we show how such an approach potentially replaces one human reviewer, enabling AI-assisted assessment of research papers. We discuss how dataset imbalances, validation procedures, and implementation time limit the broad applicability of such approaches. Through this, we develop initial guidance on the utility of proprietary LLMs in evidence synthesis.

scientific communication and education↗

bioRxiv2026

Medication and Developmental Stage Shape the Brain-Age Gap in ADHD

Identifying robust biomarkers for Attention-Deficit/Hyperactivity Disorder (ADHD) remains challenging. The brain-age gap (BAG), which is the difference between predicted biological brain age and chronological age, has been proposed as a marker of atypical neurodevelopment, yet findings have been inconsistent across studies, limiting its clinical translation. We hypothesized that two major sources of heterogeneity, medication status and developmental stage, contribute to these discrepancies. To test this, we analyzed structural MRI data from a large multi-site cohort (N = 947; ages 7-26 years). Individuals with ADHD were stratified by medication history (medicated vs. unmedicated) and developmental stage (childhood vs. adolescence) and compared with matched typically developing controls. We estimated global- and network-level BAG using publicly available pretrained models. Our results revealed that BAG differences were contingent on medication status, with abnormalities predominantly observed in unmedicated individuals. Developmental stage further moderated the presence of this effect: BAG abnormalities were present significantly in children with ADHD but not in adolescents. Network-level analyses demonstrated spatial specificity, with significant effects in visual, ventral attention, and dorsal attention networks. Together, these findings provide compelling evidence that BAG in ADHD is both medication- and development-dependent, highlighting the need to account for treatment exposure and developmental stage when evaluating BAG as a biomarker and further underscoring the importance of early intervention during childhood.

neuroscience↗

bioRxiv2025

Longitudinal in vivo MR elastography reveals whole-liver viscoelastic involvement in a murine model of hepatocellular carcinoma

Cancer cells actively shape their microenvironment and adapt the physical and biomechanical properties of the host tissue. However, for organs like the liver, especially in vivo, it remains unclear at what rate and spatial extent macroscopic viscoelastic properties change during formation of a cancer-permissive environment. Using clinical multifrequency magnetic resonance elastography (MRE) in a mouse model of hepatocellular carcinoma (HCC), we identified surprisingly early and large-scale viscoelasticity changes leading to whole-liver biomechanical involvement stretching far beyond the local cancer microenvironment. Widespread liver softening began two weeks after HCC inoculation, followed by a decrease in tissue viscosity and fluidity two weeks later, preceding any macroscopic evidence of local tumor growth. In contrast, local lesions with stiff-rigid biomechanical properties were not detectable by standard MRI and MRE until five to six weeks post-injection. Furthermore, tumor viscoelasticity correlated with that of the host liver, also suggesting a possible widespread adaptation of the biomechanical properties beyond the tumor margins and its local niche during early liver colonization. The observed large-scale viscoelastic signature, detectable just two weeks after tumor cells injection, could serve as a non-invasive imaging biomarker to inform physicians about tumor niche formation and liver cancer progression long before any macroscopic manifestation of solid tumors.

biophysics↗

bioRxiv2025

CDK12 condensation in nuclear speckles confers sensitivity to cyclin K molecular glue degraders

Transcriptional machinery often operates in the configuration of biomolecular condensates, presenting emerging opportunities to develop therapies targeting transcriptionally addicted malignancies. In this study, we discovered that cyclin-dependent kinase 12 (CDK12), a key regulator of transcription elongation, formed dynamic liquid-like droplets within nuclear speckles, a process enhanced by high CDK12 expression and CDK12 inhibitor treatment. Recognizing this unique property, we rationally designed ZSQ253, a modular cyclin K molecular glue degrader that coupled binding moieties for both CDK12 and DDB1. By promoting phase separation-driven assembly and subsequent proteasomal degradation of the CDK12-cyclin K complex, ZSQ253 interfered with oncogenic transcription and demonstrated potent antitumor capability. Further structure-guided optimization yielded HQY1428, a derivative with a multi-site gluing effect and improved drug activity. Given the frequent genomic co-amplification of CDK12 and ERBB2 across diverse human cancers, we showed that combining HQY1428 with HER2 inhibitors provided synergistic therapeutic benefits. Collectively, our findings establish CDK12 condensation within nuclear speckles as an exploitable vulnerability, and introduce a novel mechanism of action underlying cyclin K molecular glue degraders.

cancer biology↗

bioRxiv2025

DiCE: Differential Centrality-Ensemble Analysis Based on Gene Expression Profiles and Protein-Protein Interaction Network

Uncovering key genes that drive diseases and cancers is crucial for advancing understanding and developing targeted therapies. Traditional differential expression analysis often relies on arbitrary cutoffs, missing critical genes with subtle expression changes. Some methods incorporate protein-protein interactions (PPIs) but depend on prior disease knowledge. To address these challenges, we developed DiCE (Differential Centrality-Ensemble), a novel approach that combines differential expression with network centrality analysis, independent of prior disease annotations. DiCE identifies candidate genes, refines them with an information gain filter, and reconstructs a condition-specific weighted PPI network. Using centrality measures, DiCE ranks genes based on expression shifts and network influence. Validated on prostate cancer datasets, DiCE identified genes over-represented in key pathways and cancer fitness genes, significantly correlating with disease-free survival (DFS), despite DFS not being used in selection. DiCE offers a comprehensive, unbiased approach to identifying disease-associated genes, advancing biomarker discovery and therapeutic development. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=77 SRC="FIGDIR/small/638654v1_ufig1.gif" ALT="Figure 1"> View larger version (15K): org.highwire.dtl.DTLVardef@1ce9b81org.highwire.dtl.DTLVardef@14352e1org.highwire.dtl.DTLVardef@f63caeorg.highwire.dtl.DTLVardef@a0086_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2025

clusIBD: Robust Detection of Identity-by-descent (IBD) Segments Using Unphased Genetic Data from Poor-quality DNA Samples

Inferring relatedness through the detection of identity-by-descent (IBD) segments is widely used in many fields. However, existing methods are often powerless for poor-quality DNA samples. Here, we propose a method, clusIBD, which can robustly detect IBD segments using unphased genetic data with high genotype errors. We evaluated and compared the performance of clusIBD with IBIS, TRUFFLE, and IBDseq using simulated data, artificial poor-quality materials and ancient DNA samples. The results showed that clusIBD outperformed these state-of-the-art existing tools and may be a promising tool for kinship inference in fields such as ancient DNA analysis and criminal investigation (freely available at https://github.com/Ryan620/clusIBD/).

genetics↗

bioRxiv2025

Chimeric Collagen-like Proteins with Tunable Structural Heterogeneity for Precise Control and Targeted Doxorubicin Delivery

Protein-based drug delivery systems offer advantages such as genetic tunability, structural homogeneity, and high biocompatibility. However, precisely controlling the assembly of protein nanoparticles and establishing correlations between their macroscopic properties and molecular architecture remain significant challenges. Here, inspired by chimeric collagen-like proteins, we rationally designed a series of de novo chimeric proteins comprising an N-terminal globular domain fused to a linear collagen-like domain. These structurally heterogeneous proteins were conjugated with doxorubicin (DOX) and homogenized to form chimeric collagen-like protein-DOX nanocomplexes (CDCNs). By fine-tuning the protein structure and domain organization, we achieved precise control over CDCN architecture and drug release kinetics. Furthermore, to enhance tumor specificity, triple-negative breast cancer (TNBC)- and glioblastoma (GBM)-targeting peptides were genetically fused to the chimeric proteins, respectively. In both in vitro and in vivo models of TNBC and GBM, CDCNs facilitated selective tumor accumulation, enhanced cellular uptake, and promoted apoptosis while minimizing off-target toxicity. This work establishes a strategy for designing protein-based nanoplatforms with programmable structures and tunable functionalities, offering promising potential for precisely controlling of protein-based delivery system.

bioengineering↗

bioRxiv2024

D-lactate derived from intestinal bacteria drives lysine D-lactylation to modulate transcription in liver cells

Emerging evidence indicates that the accumulated D-lactate (D-la) from intestinal bacteria induces diverse effects in human intestinal injury diseases. However, its regulatory mechanism and functional consequence remain unclear. Here, we show that the secreted D-la induces lysine D-lactylation (K-Dla) to modulate transcription in host cells. We uncover K-Dla in HepG2 cells stimulated by D-la from E. coli secretions. Then, we confirm the occurrence of D-la-driven K-Dla by four orthogonal approaches. We further validate that the existence of K-Dla in organisms by detecting the mouse liver and human intestinal tissue. Besides, we demonstrate a SCOT1-catalyzed biosynthetic pathway of D-lactyl-CoA, serving as a key donor for K-Dla. Next, we identify the K-Dla landscape with 2895 sites and further suggest the regulation of histone K-Dla in transcription by MNase ChIP-seq and RNA-seq assays. Finally, we show the potential effect of K-Dla on inhibiting the hepatocellular carcinoma cell proliferation. In summary, we report that the secreted D-la can induce K-Dla to regulate transcription, providing a new insight into the function of D-la secreted by gut microbiota and the effect of protein chemical modifications induced by bacterial secretions.

molecular biology↗

bioRxiv2024

Gradients of Recognition Molecules Shape Synaptic Specificity of a Visuomotor Transformation

Converting sensory information into motor commands is fundamental to most of our actions. In Drosophila, visuomotor transformations are mediated by Visual Projection Neurons (VPNs). These neurons convert object location and motion into directional behaviors downstream through a synaptic gradient mechanism. However, the molecular origins of such graded connectivity remain unknown. We addressed this question in a VPN cell type called LPLC2, which integrates looming motion and transforms it into an escape response through two parallel dorsoventral synaptic gradients at its inputs and outputs. We identified two corresponding dorsoventral expression gradients of cell recognition molecules within the LPLC2 population that regulate this synaptic connectivity. Dpr13 determines synaptic outputs of LPLC2 axons by interacting with its binding partner DIP-{varepsilon} expressed in the Giant Fiber, a neuron that mediates escape. Similarly, beat-VI regulates synaptic inputs onto LPLC2 dendrites by interacting with Side-II expressed in upstream motion-detecting neurons. Behavioral, physiological, and molecular experiments demonstrate that these coordinated molecular gradients control differential synaptic connectivity, enabling the accurate transformation of visual features into motor commands. As within-neuronal-type variation in gene expression is also observed in the mammalian brain, graded expression of cell recognition molecules may represent a common mechanism underlying synaptic specificity.

neuroscience↗