Biology subjects

Wu, N.

Publications and source records attributed to Wu, N..

At least 19 recordsLinked to original sources

bioRxiv2026

Goal directed closed-loop training of network bursts in cortical cultures produces improved performance and error correction over days

Dissociated neural cultures are living systems with no evolutionary history in their current form, which makes them a testbed for asking what goals neural tissue can learn to pursue once it is removed from its evolved body. It remains unclear which features of network activity such a culture can be trained to control, and whether that training is retained from one day to the next. We built an open-source closed-loop platform, the Holistic Agential Learning platform (HAL), that detects a network burst on a high-density microelectrode array, estimates its spatial origin in real time, and applies a minimal reinforcement rule. Here, a burst that originates off the goal side triggers global random stimulation that persists until the next burst. Over days of training, rat cortical cultures (n=10) increased the fraction of bursts that initiated on the goal side. This gain accrued between sessions, with little change within a session, a consolidation-like pattern that we interpret as a statement about timescale while the mechanism remains untested. The burst origin moved toward the goal side, while the burst peak and the repertoire of propagation directions were preserved. Trained cultures also recovered from incorrect states more reliably over days, while the rate at which they entered those states was unchanged, and this recovery effect was the most robust difference from unstimulated controls (n=6). Cultures that began on either side of chance converged to a similar level of performance. Damaged cultures (n=6) with comparable burst statistics showed neither the across-day gain nor the consolidation-like pattern. Training reweighted the activity patterns each culture already had, and we saw no new ones, which we read as a simple form of niche construction in which the tissue keeps a trace of its history that biases what it does next. These results identify burst initiation and error recovery as trainable variables in a novel living system, and they support training the tissue rather than engineering the machine.

bioengineering↗

bioRxiv2026

DEPICT enables sample-centric therapeutic prioritization across evolving and spatially heterogeneous tumor states

Accurate treatment selection remains a major challenge in precision oncology because tumors with similar genomic alterations can respond differently to therapy. Most drug-response models evaluate individual drugs or drug-sample pairs, whereas clinical decision-making requires therapies to be prioritized within each tumor context. Here we present DEPICT (Drug Efficacy Prediction via Integrated ConText), a sample-centric deep learning framework that reframes drug-response prediction as the relative prioritization of therapies within an individual tumor context. DEPICT integrates transcriptomic and mutational features with prior alteration-therapy relationship and was trained on 231,880 cell line-drug pairs comprising 941 cancer cell lines and 289 compounds. Without retraining, DEPICT retained predictive performance in independently profiled cell lines, experimentally induced drug-resistant states and matched primary lung tumors and patient-derived organoids. Representations learned exclusively from bulk pharmacogenomic data further captured progressive drug-response states at single-cell resolution and outperformed existing single-cell drug-response inference methods. Applied to spatial transcriptomic data, DEPICT resolved clone-associated therapeutic heterogeneity within individual lung tumors, and predicted regional differences were recapitulated in matched multiregional organoid drug-response assays. These findings establish sample-centric therapeutic prioritization as a transferable strategy for modeling dynamic and spatially heterogeneous drug vulnerabilities across biological scales.

bioinformatics↗

bioRxiv2026

EvSpark: Lossless Speculative Decoding for Hybrid DNA Foundation Models

Hybrid DNA foundation models combine convolutional, recurrent, and attention layers, making speculative decoding more difficult than truncating a KV cache. We present EvSpark, a speculative decoding system for Evo2 that verifies draft blocks in parallel and restores all three classes of inference state by selecting retained intermediate states, without replay. A compact, hidden-state-conditioned drafter proposes each block in one parallel forward pass. On Evo2 7B, a 48-prompt benchmark with three training seeds yields 2.96x on 43 real-sequence prompts and 3.27x including the five synthetic controls. Acceleration persists at 262k-token context (1.84x - 2.43x on two bacterial genomes) and over 32k generated tokens. Retraining the same drafter architecture for Evo2 20B and 40B yields 2.18x - 2.46x on real sequences and 2.51x - 2.78x on the full suite. Autoregressive drafter comparisons and batch measurements show why low draft latency, rather than acceptance alone, determines the gain. The method preserves the target distribution in exact arithmetic. In bf16, greedy tests find no non-tie divergences across 48 prompts and six checkpoints; sampling tests expose residual numerical sensitivity, especially in repetitive sequences. A cost-efficient 7B drafter requires 1.06 incremental GPU-hours of training, excluding teacher-data collection, and achieves 2.82x on real sequences. In regulatory-DNA design, EvSpark achieves a median complete-workflow speedup of 1.57x over a calibrated batched native baseline.

bioinformatics↗

bioRxiv2026

Avs2 drives non-canonical tetrameric assembly of trypsin-like domain for anti-phage defense

Bacteria have evolved diverse anti-phage defense systems, with the antiviral STAND family (Avs) representing one of the most diverse and widespread, encompassing at least 90 distinct families, yet only Avs3, Avs4, Avs5 and Avs7 have been well characterized. Here, we elucidated the molecular mechanism of Avs2-trypsin-MBL system, where phage terminase recognition by Avs2 triggers coupled activation of the protease and nuclease activities of trypsin-MBL. Cryo-EM structure of Avs2-trypsin-terminase and biochemical analysis reveal that the binding of terminase ATPase domain triggers the assembly of Avs2 into tetramer, with two unique ATP molecules bridging ATPase active-site recognition by TPR domain. This tetramerization drives the fused trypsin into an active C4-symmetric assembly, an architecture distinct from the conventional non-defense trypsin. Our study unravels the activation mechanism of Avs2-trypsin-MBL system, expanding our understanding on commonality and diversity of widespread Avs-mediated anti-phage immunity, alongside the structural and functional adaption of trypsin.

microbiology↗

bioRxiv2026

Physical laws predict methane hotspots in global mountain waters

Accurate accounting of aquatic methane emissions is critical for climate change mitigation, yet current global budgets overlook a key driver: the elevation-regulated atmospheric pressure. Here, we present the first large-scale investigation of methane ebullition across 164 shallow waters spanning elevations from sea level to 4886 meters. We demonstrate that ebullition rate increases with elevation-over four times higher at >3000 m a.s.l. than at sea level-due to two synergistic, pressure-dependent physical mechanisms: a degas effect (enhanced bubble formation) and a trigger effect (facilitated bubble ascent). Independent theoretical prediction of the combined effects shows near-perfect agreement with the empirical elevation trend, quantitatively confirming that these physical mechanisms are the primary drivers of enhanced ebullition at high elevations. Our findings reveal that mountain aquatic ecosystems represent unaccounted methane hotspots that have been systematically underestimated in global inventories.We therefore call for urgent integration of these ecosystems into IPCC assessments and targeted mountain mitigation and sustainable management strategies.

ecology↗

bioRxiv2026

Systematic In Silico Off-Target Assessment of siRNAs: Integrated Tissue-Specific Scoring and Cross-Species Preclinical Model Selection with TargetSureR

Small interfering RNAs (siRNAs) have become a transformative class of nucleic acid therapeutics for clinical disease treatment, yet sequence-dependent off-target silencing continues to pose a major safety barrier that hinders their preclinical refinement and large-scale translational application. Existing bioinformatics tools only support partial off-target evaluation, either focusing on basic sequence optimization or simple seed-region scanning, and fail to deliver systematic, multi-dimensional and reproducible safety assessment for siRNA lead screening. To fill this gap, we developed TargetSureR, a lightweight, modular and CRAN-compatible open-source R package dedicated to full-process siRNA off-target risk profiling. This tool accommodates dual sequence-based and precomputed position-based inputs, integrates GTEx multi-tissue expression data and curated cancer, adverse-event and immune gene panels, and establishes a seven-dimensional scoring framework to stratify off-target risks into four hierarchical tiers. It further enables tissue-specific safety characterization and quantitative cross-species model selection, with an Ensembl API fallback mechanism ensuring high transcript annotation resolution. Built purely in R with no external shell dependencies, TargetSureR provides a standardized, robust and user-friendly workflow for systematic siRNA preclinical safety evaluation, and is freely available at https://github.com/nishuai/TargetSureR.

bioinformatics↗

bioRxiv2026

Prediction of Transcription Factor DNA Binding Affinity with High-Throughput Kd Measurements and Deep Learning

Transcription factors (TFs) regulate gene expression through specific interactions with genomic DNA. While TF binding motifs from public databases describe sequence preferences, quantifying genome-wide affinity (Kd) is highly desirable for a more accurate thermodynamic description. Here, we report ivtFOODIE (in vitro FOOtprinting with DeamInasE), an assay that leverages deaminase-mediated cytosine-to-uracil conversion to measure Kd values for a given TF across accessible genomic regions from human cells. By pre-training on TF binding sites from JASPAR and fine-tuning with our ivtFOODIE data from 46 TFs representing 13 different DNA-binding domains (DBDs), we developed Seq2Kd, a deep learning model capable of predicting a TFs absolute binding affinity on DNA sequences. Seq2Kd enables de novo motif discovery of [~]500 previously uncharacterized human TFs and reveals the effects of genetic variation both in TF-coding regions and DNA-binding sites on gene expression and disease susceptibility. By correlating predicted affinity changes with the sign and magnitude of expression quantitative trait locus (eQTL) effects, we stratified TFs into activator-like and repressor-like groups. Compared to clinically benign variants, pathogenic single-nucleotide variants (SNVs) within regulatory and protein-coding regions show significantly larger predicted shifts in Kd. We provide an interactive web portal, the ENcyclopedia of Transcription-factor Interactions with Regulatory Elements (ENTIRE), which integrates the Seq2Kd model with the ivtFOODIE dataset. This resource offers thermodynamic prediction for TF-DNA interactions for functional genomics and human disease.

biophysics↗

bioRxiv2026

The FERONIA receptor kinase is required for high humidity responses in Arabidopsis

High humidity greatly influences plant growth and development and triggers adaptive physiological responses such as leaf hyponasty (elongation of leaf petiole and upward leaf movement). A recent study identified Cyclic Nucleotide-Gated Ion Channels 2 and 4 (CNGC2/4)-mediated Ca2+ influx and Calmodulin Binding Transcription Activators 2 and 3 (CAMTA2/3)-mediated transcription as essential for high humidity response in Arabidopsis, but the upstream regulators that control these pathways remain unknown. Here, we show that the receptor-like kinase FERONIA and its co-receptor LORELEI-LIKE GPI-ANCHORED PROTEIN1 (LLG1) are required for a large portion of high humidity-associated Arabidopsis transcriptomic changes, including CNGC2, CAMTA-regulated genes, and cell wall remodeling genes, and for high humidity-induced leaf hyponasty. High humidity triggers a previously uncharacterized petiole-localized Ca2+ waves that precede hyponastic leaf movement. The petiole-localized Ca2+ signals were significantly altered in the fer-4 mutant. Thus, FERONIA is a key regulator of plant responses to extracellular high humidity. Highlights FERONIA plays a prominent role in transcriptomic responses to high humidity FERONIA is required for high humidity-induced leaf hyponasty High humidity induces petiole calcium waves FERONIA is required for normal petiole calcium waves in response to high humidity

plant biology↗

bioRxiv2026

Metabolic dehalogenation of halomethanes by mercury methylators

Mercury (Hg) methylators play key roles in the global Hg cycle. Nonetheless, the Hg methylators are widely present in Hg-absent environments and the Hg methylation is a cometabolic process without cell growth, which leave the origin and evolution of the Hg methylation as a mystery. Here, we reported the CCl4/trihalomethane-to-dihalomethane dehalogenation by model Hg methylators in a metabolic way. Heterologously-expressed HgcAB catalyzed both the CCl4 dechlorination and Hg methylation. Halomethanes were shown to sustain the Hg-methylation community without adding external carbon source, electron donor and acceptor. Metadata analyses suggested the halomethane dehalogenation potential of Hg methylators at the global scale. These results, together with much higher global flux and kinetic Vmax of the metabolic halomethanes dehalogenation relative to the co-metabolic Hg methylation, suggested halomethanes as the potential HgcAB substrates for the evolution of Hg methylation.

microbiology↗

bioRxiv2026

A correlational study of ABCA3 and SCN4B as exercise-related biomarkers of patients with Stanford type A aortic dissection

BackgroundAccumulating evidence indicates that moderate exercise may reduce the incidence of Stanford type A aortic dissection (TAAD), but the specific mechanisms remain unclear. This study aims to identify exercise-related biomarkers in TAAD patients and to investigate their underlying mechanisms. MethodsTranscriptome data related to TAAD and exercise-related genes were obtained from publicly available databases. Candidate biomarkers for TAAD were identified through an integrative approach incorporating differential expression analysis, machine learning, and expression level assessment, leading to the construction of a diagnostic model. Subsequently, functional enrichment, immune infiltration, regulatory network analysis, and computational drug prediction were conducted to systematically investigate the pathological mechanisms and translational potential of the indentified biomarkers. ResultsABCA3 and SCN4B were identified as exercise-related biomarkers in TAAD progression. A nomogram incorporating these two biomarkers exhibited strong diagnostic performance for identifying the disease. Functional enrichment analysis revealed potential involvement of these biomarkers in disease progression through pathways including circadian rhythm regulation and ribosome biosynthesis. Additionally, immune cells like M1 macrophages and naive B cells, as well as regulatory factors including hsa-miR-1343-3p and XIST, were found to be involved in this process. Finally, zonisamide and MRS1097 were identified through computation prediction as potential therapeutic drugs. ConclusionABCA3 and SCN4B were identified as exercise-related biomarkers associatied with TAAD and represent potential valuable targets for both diagnosis and treatment strategies.

bioinformatics↗

bioRxiv2026

CESAR: High-Sensitivity Detection of Copy Number Variations in ctDNA Using Segmentation and Anchor Recalibration

BackgroundDetecting copy number variations (CNVs) in circulating tumor DNA (ctDNA) is crucial for the companion diagnosis and resistance monitoring of various solid tumors (e.g., NSCLC, Glioblastoma). However, when tumor-derived DNA fractions are extremely low (often <1%), traditional depth-based methods frequently fail due to non-linear sequencing depth fluctuations and probe-specific capture biases inherent to targeted Next-Generation Sequencing (NGS). MethodsWe developed CESAR (CNV Estimation with Segmentation and Anchor Recalibration), a computational tool for tumor-only CNV detection in targeted NGS panels. CESAR uses Circular Binary Segmentation (CBS) to re-partition target regions according to relative capture efficiency, then applies a data-driven "anchor" selection procedure that, for each target segment, identifies a personalized set of co-varying genomic segments. By selecting the anchor set that minimizes the coefficient of variation (CV) of the anchor-recalibrated depth ratio across a panel of normals, CESAR recalibrates the per-segment baseline and suppresses probe-specific technical noise. Copy-number status is then called from the deviation of the observed ratio against this trained baseline. ResultsUsing standard DNA reference materials, CESAR identified amplifications of MET, ERBB2, and EGFR at low tumor fractions. CESAR resolved focal alterations as subtle as 2.18 copies (a 1.09-fold change relative to the diploid baseline) while reporting no false-positive amplifications in diploid control regions. Applied to a clinical cohort of nine NSCLC ctDNA samples profiled on a 94-gene panel, CESAR successfully separated three MET-amplified samples from six matched negatives, where the standard CNVkit-based pipeline failed to distinguish them. In head-to-head benchmarking on identical data, CESAR reduced technical variance relative to the widely used CNVkit and more reproducibly resolved low-level copy-number gains, particularly in the depth-heterogeneous MET amplicon. ConclusionsCESAR provides a stable and sensitive framework for tumor-only CNV calling in liquid biopsies. On reference standards it outperforms CNVkit in bias and reproducibility, and on a clinical NSCLC cohort it recovered MET CNV abnormalities that a standard CNVkit pipeline missed. Validation on larger and more diverse cohorts is warranted.

bioinformatics↗

bioRxiv2026

Cystinosin/Ers1 functions in redox homeostasis in the early secretory pathway

Cystinosis is an autosomal recessive inherited disorder caused by mutations in the CTNS gene, which encodes the highly conserved transmembrane protein cystinosin, a proton/cystine co-transporter at the lysosome membrane. However, reduction of cystine load in the lysosomes is insufficient to treat key disease symptoms, indicating that cystinosin performs additional disease-relevant functions. Here, we report that Ers1, the yeast homolog of cystinosin, localizes to and functions in the early secretory pathway. We provide evidence that Ers1 does not transport cystine. Ers1 genetically interacts with early secretory pathway recycling adaptors and redox-active Fe-S cluster-binding proteins. Notably, cystinosin-LKG, the extra-lysosomal localized splicing isoform of cystinosin, can functionally replace Ers1 in yeast. Collectively, our work uncovers a conserved role of cystinosin/Ers1 in the early secretory pathway, offering new molecular insights for understanding cystinosis pathology.

cell biology↗

bioRxiv2025

LISA: A Comprehensive R Package for Lentiviral Integration Site Analysis in Gene Therapy Safety Assessment

BackgroundGene therapy using integrating viral vectors has shown remarkable clinical success, but the random integration of vectors into the host genome poses potential safety concerns, particularly insertional mutagenesis. Common Integration Sites (CIS) near oncogenes significantly increase the risk of clonal expansion and malignant transformation. Current tools for integration site (IS) analysis lack comprehensive functionality for safety assessment, particularly in identifying risk-associated genomic regions and tracking clonal dynamics over time. ResultsWe developed LISA (Longitudinal Integration Site Analysis), a comprehensive R package specifically designed for gene therapy safety evaluation. LISA incorporates a novel approach that trains integration site profiles using cumulative curves derived from normal cell infusion samples and normal patient samples, with subsequent statistical testing to distinguish these normal samples from those with exceptional clonal expansion, thereby enabling effective identification of dominant clone formation. For longitudinal studies, LISA monitors fluctuations in clonal dynamics via Richness and Evenness metrics, and integrates this cumulative curve-based statistical testing approach into its early warning system to detect potentially dominant clones. Additionally, LISA features an optimized CIS detection algorithm with O (n log n) time complexity, achieved through chromosome-position encoding and a seed-site strategy, which facilitates efficient analysis of large-scale clinical datasets. The package also provides systematic functional annotations, including annotations for Safe Harbor regions, enhancer/promoter elements, and proximity to three curated gene sets: adverse event-associated genes (AE genes, n=36), cancer-related genes (n=980), and immune-related genes (n=2,485). Furthermore, it encompasses six categories of interactive visualizations and has been validated using real clinical data from gene therapy trials.

bioinformatics↗

bioRxiv2025

The anti-phage mechanism of a widespread trypsin-MBL module

Protease mediate activation of immune effectors is a conserved mechanism across life. This study identified a widespread and co-evolving trypsin-MBL module as a core effector in diverse bacterial immune systems (e.g., Hachiman, Argonaute, AVAST), protecting host via abortive infection. Focusing on Hachiman-trypsin-MBL, we demonstrated that, the protease activity of trypsin*HamAB is inhibited in the presence of ATP and MBL is an autoinhibited DNase with two inserted loops blocking the catalytic pocket. Upon infection, the HamB helicase senses invading nucleic acids and catalyzes the hydrolysis of ATP, activating the associated trypsin-like protease that cleaves specifically at two non-canonical KSS motif of MBL and releases its DNase activity. Cryo-EM structures of activated trypsin*HamAB bound with DNA reveal DNA binding and ATP hydrolysis lead to the release of the trypsin-like domain, enabling activation. Our work provides a comprehensive understanding on the mechanism of widely distributed trypsin-MBL module, wherein proteolytic activation of a toxic nuclease enables robust bacterial immunity while preventing self-toxicity through multi-layered control.

biochemistry↗

bioRxiv2025

Segmentation Matters: Recognizing the Cell Segmentation Challenge in Spatial Transcriptomics

Probe-based in situ hybridization spatial transcriptomics has emerged as a state-of-the-art for neuroscience research. Accurate segmentation of neurons and non-neuronal cells, a critical step for downstream analysis, remains a big challenge. Using human sensory ganglia neurons as an example, we systematically explore this problem. We evaluated multiple automatic segmentation approaches using quantitative performance metrics and downstream results. We show that careful parameter tuning is essential for achieving accurate segmentation; however, even with optimized parameters, different models still yield distinct classes of errors. To mitigate these errors, we propose a manual quality check, which can validate and refine automated segmentation results. As a future direction, integrating multi-modal imaging data with tailored neural networks may further improve segmentation accuracy. In summary, our results argue that each automated segmentation method has distinct strengths, weaknesses, and characteristic error patterns; and that a manual review is necessary.

neuroscience↗

bioRxiv2025

c-Rel drives pancreatic cancer metastasis through Fibronectin-Integrin signaling-induced isolation stress resistance and EMT activation

Pancreatic ductal adenocarcinoma remains one of the deadliest malignancies, with limited treatment options and a high recurrence rate. Recurrence happens often with metastasis, for which cancer cells must adapt to isolation stress to successfully colonize distant organs. While the fibronectin-integrin axis has been implicated in this adaptation, its regulatory mechanisms require further elaboration. Here, we identify c-Rel as an oncogenic driver in PDAC, promoting epithelial-to-mesenchymal transition (EMT) plasticity, extracellular matrix (ECM) remodeling, and resistance to isolation stress. Mechanistically, c-Rel directly regulates fibronectin (Fn1) and CD61 (itgb3) transcription, enhancing cellular plasticity and survival under anchorage-independent conditions. Fibronectin is not essential for EMT, but its absence significantly impairs metastatic colonization, highlighting a tumor-autonomous role for FN1 in isolation stress adaptation. These findings establish c-Rel as a key regulator of PDAC metastasis by controlling circulating tumor cell (CTC) niche and survival, suggesting that targeting the c-Rel-fibronectin-integrin axis could provide new therapeutic strategies to mitigate disease progression and recurrence.

cancer biology↗

bioRxiv2024

Tachykinin1-expressing neurons in the parasubthalamic nucleus control active avoidance learning

Active avoidance is a type of instrumental behavior that requires an organism actively to engage in specific actions to avoid or escape from a potentially aversive stimulus and is crucial for the survival and well-being of organisms. It requires a widely distributed, hard-wired neural circuits spanning multiple brain regions, including the amygdala and thalamus. However, less is known about whether and how the hypothalamus encodes and controls active avoidance learning. Here we identify a previously unknown role for the parasubthalamic nucleus (PSTN), located in the lateral subdivision of the posterior hypothalamus, in the encoding and control of active avoidance learning. Fiber photometry calcium imaging shows that the activity of tachykinin1-expressing PSTN (PSTNTac1) neurons progressively increases during this learning. Cell-type specific ablation and optogenetic inhibition of PSTNTac1 neurons attenuates active avoidance learning, whereas optogenetic activation of these cells promotes this learning via a negative motivational drive. Moreover, the PSTN mediates this learning differentially through its downstream targets. Together, this study identifies the PSTN as a new member of the neural networks involved in active avoidance learning and offers us potential implications for therapeutic interventions targeting anxiety disorders and other conditions involving maladaptive avoidance learning.

neuroscience↗

bioRxiv2024

Embryonic Mast Cells Arise from Cpa3-expressing Precursors Independent of Granulocyte-Monocyte Progenitors

Most of the mast cells (MCs) in connective tissues, such as skin, are long-lived embryonic-derived immune cells that play important roles in host defense and various immunological diseases, including allergies. Their embryonic origin and ontogeny remain to be fully studied since several overlapping waves of embryonic hematopoiesis have been shown to give rise to these cells. Here, combining single-cell RNA sequencing and new genetic fate mapping models, we identified a Cpa3-expressing population sequentially appearing in the yolk sac, fetal liver, and peripheral tissues which gives rise to dermal MCs during embryonic days 11.5 to 14.5. Using in vitro differentiation and in vivo transplant methods, we identified a Ter119-F4/80- CD45+CD117+CD16/32+CD135-CD115-Ly6C-CD34lo population as potential fetal liver MC precursors. Fate mapping with Cpa3CreERT2and Zbtb16CreERT2 models, as well as the granulocyte-monocyte progenitor (GMP) models Ms4a3Cre and ElaneCre, demonstrated that MCs arise from Cpa3+ progenitors rather than Ms4a3+ or Elane+ GMPs. A corresponding population with a similar developmental trajectory was also identified in human early yolk sac and fetal liver, suggesting a conserved MC developmental program across species. These findings delineate the detailed developmental path of MCs in embryos, permitting future functional studies in immunity and development. HighlightsO_LICpa3-expressing cells in the yolk sac and fetal liver contain mast cell precursors C_LIO_LICpa3CreERT2 labels embryonic mast cells and their precursors C_LIO_LIEmbryonic mast cells arise through Cpa3+ mast cell precursors, but not Elane+/Ms4a3+ GMPs C_LIO_LIThe seeding of MCPs to embryonic skin slows down after E14.5 C_LI

immunology↗