Biology subjects

Lu, D.

Publications and source records attributed to Lu, D..

At least 19 recordsLinked to original sources

bioRxiv2026

EpiRank: partner-agnostic surface-patch retrieval for epitope prediction of protein binding interfaces

Structure-based interface predictors are normally reported as a single pooled ROC AUC over a residue test set. For a protein designer the operative question is different and much harder: given one target chain, will the top-ranked surface patch be a real binding site? We built EpiRank, a partner-agnostic epitope predictor, on HumanPPI 29,246 predicted human dimers, 12,296 target chains, 4.18 M residues using only single-chain geometry and physicochemistry (170 features, no partner, no sequence search, no structural alphabet), and evaluated it as a retrieval problem with homology- and partner-graph-controlled splits, on the 3,529 chains that survive an explicit partner-contamination filter. The model places a real epitope in its top-ranked patch 64 % of the time. That number is far less informative than it looks. Compact patches placed at random hit a real epitope 41 % of the time under the same rule, so the model's lift is 1.57 x, not 64 %; 54 % of all predicted patches land on nothing; and when the same chains are scored by models trained on a 2,000-chain and an 11,669-chain superset, a substantial fraction change answer while the aggregate barely moves. We document four specific ways this task invites overstatement partner-graph percolation across splits, single-linkage homology percolation at the conventional 30 % identity threshold, a count-only hit rule that rewards spatial smoothing rather than signal, and hotspot precision conditioned on success, which ranks a random baseline above the trained model. Each was hit during development and each inverted a conclusion we had already drawn. We argue that interface-prediction papers should report lift over a geometry-matched random baseline, patch precision, and per-target stability alongside AUC, and we release the evaluation harness that computes them.

bioinformatics↗

bioRxiv2026

Germinal centers can persist for up to one year with declining somatic hypermutation and ongoing plasma cell output

The total duration of primary germinal center (GC) responses, and how B cell evolution changes as these responses mature, remain poorly defined. Here, longitudinal sampling of rhesus macaques immunized with HIV Envelope protein revealed that a single slow-delivery prime sustained ongoing GCs for nearly a year without additional antigen. Somatic hypermutation (SHM) accumulated throughout these prolonged responses, but its rate declined substantially over time, independent of both priming strategy and boosting. Over the same period, subdominant non-base Env specificities became increasingly represented within the GC response. Delaying boosting until after primary GC contraction increased recruitment of recalled memory B cells into the GC response. Bone marrow plasma cells were generated throughout prolonged GC maturation, with the timing of their generation modulated by booster timing. Together, these findings reveal how GCs evolve across nearly a year and how this prolonged maturation shapes the memory and plasma cell compartments that underpin long-term humoral immunity.

immunology↗

bioRxiv2026

A closed-loop reinforcement learning framework for rapid compound directed optimization

Generative artificial intelligence (AI) holds transformative potential for drug discovery, yet existing architectures typically operate in open loops without experimental feedback. Here we introduce rapid compound directed optimization (RCDO), a closed-loop reinforcement learning framework that accelerates the optimization process by bridging dry-lab computation with wet-lab feedback. RCDO couples a three-dimensional structure-guided generative model with a multi-level reward system updated after each design cycle using experimental measurements from all synthesized compounds, including inactive or developability-failed compounds. By continuously aligning the generative model with accumulated wet-lab measurements, RCDO substantially compresses optimization timelines. We evaluated RCDO through retrospective benchmarking against historical optimization trajectories and prospective wet-lab campaigns targeting ROR1, NLRP3, and NSD3. Across prospective evaluations, RCDO rapidly resolved key optimization bottlenecks within two to three design cycles: improving the oral exposure of an ROR1 inhibitor by 40-fold while maintaining antitumor efficacy, reducing CYP2C19 inhibition of an NLRP3 antagonist by 20-fold while preserving inflammasome activity, and boosting the binding affinity of an NSD3 hit by 18-fold. By directly coupling wet-lab feedback to generative learning, RCDO establishes an efficient platform for compound directed optimization, transforming AI-driven drug discovery from static generation into continuous experimental adaptation.

bioinformatics↗

bioRxiv2026

rVSV-EBOV vaccination protects ferrets from lethal Bundibugyo virus disease

An uncontrolled and rapidly growing outbreak of Bundibugyo virus (BDBV) is currently gripping the Democratic Republic of the Congo and threatening health security across Central Africa. There are no available BDBV-specific vaccines, although emerging evidence suggests that the Ebola virus-specific vaccine, rVSV-EBOV (also known by its tradename ERVEBO), may offer cross-protective immunity. To directly address this question, we evaluated the efficacy of rVSV-EBOV in the uniformly lethal ferret model of BDBV infection. All vaccinated animals survived BDBV challenge and exhibited minimal clinical signs of infection, presumably as a result of a moderate--but protective--humoral immune response. These findings provide critical evidence further supporting the cross-protective efficacy of rVSV-EBOV, and they suggest a potential role for this vaccine in mitigating the ongoing BDBV outbreak.

microbiology↗

bioRxiv2026

Virtual spatial transcriptomics from histopathology enables prognostic and therapeutic response prediction in cancer

Spatial transcriptomics reveals cellular heterogeneity, intercellular communication, and tissue organization, but its cost and limited accessibility restrict clinical use. Here, we present VISTA, a model that integrates multi-scale histological features and spatial context to infer spatial gene expression from H&E-stained tissue images. Across leave-one-section-out cross-validation and independent validation, VISTA robustly predicted thousands of genes and outperformed state-of-the-art methods. Beyond expression reconstruction, VISTA enabled clinically relevant downstream analyses. In TCGA breast cancer samples, it identified survival-associated genes, stratified prognostic risk groups, and revealed adverse tumor-associated spatial subtypes. In our in-house intrahepatic cholangiocarcinoma cohort, it preserved tumor-normal organization and identified CLDN4 and CYP3A4 as complementary spatial biomarkers. In HER2+ breast cancer, it predicted pathological response to neoadjuvant trastuzumab-based therapy and linked response-associated regions to immune and cytokine-related programs. These results support virtual spatial transcriptomics from routine histopathology for oncology applications.

genomics↗

bioRxiv2026

DNA-Programmed Condensate-Membrane Wetting and Cellular Internalization

Deformable condensates offer dynamic interfaces for biomolecule delivery, yet membrane adhesion does not necessarily lead to cellular internalization. The physical transition that determines whether a membrane-bound soft material remains surface-anchored or proceeds through wetting towards productive uptake remains poorly understood, particularly at active living-cell membranes. Here, we engineer sequence-defined DNA condensates through liquid-liquid phase separation (LLPS) and program their interfacial behavior by tuning sticky-end valency and cholesterol organization. These molecular designs precisely regulate condensate fluidity, fusion dynamics and internal organization, generating distinct states of weak contact, persistent anchoring and rapid wetting. Increasing cholesterol-mediated affinity does not enhance uptake. Instead, productive internalization emerges from a balance between membrane adhesion and condensate fluidity and deformability. Native membrane composition further modulates condensate interfacial fate across mammalian cells and plant protoplasts. DNA condensates enrich and deliver CpG ODNs, mRNA ([~]2000 nt) and proteins, while cargo loading experiments reveal that preserving condensate architecture is essential for functional delivery. Our findings identify wetting competence as a design parameter for controlling soft material engagement and cellular entry.

biophysics↗

bioRxiv2026

Lineage Memory and Convergent Adaptation Dictate the Single-Cell and Spatial Architecture of Brain Malignancies

How diverse cancer lineages navigate the draconian central nervous system environment--whether constrained by ancestral ontogeny or driven by convergent adaptation--remains a fundamental biological paradox. To decode this, we integrated extensive in-house and public resources to construct the largest single-cell atlas of primary and secondary brain malignancies (>550,000 cells). We demonstrate that malignant cells balance strict lineage imprinting with shared brain-adaptive programs, powerfully converging upon a dominant pan-cancer mesenchymal (MES)-like state alongside a SYT1+ pioneer subpopulation exploiting neuronal mimicry. Concurrently, the microenvironment undergoes lineage-constrained divergence. The myeloid landscape shifts from a resident-dominated architecture in gliomas to extensive blood-borne infiltration in metastases. Lymphoid responses parallel this, crowning laryngocarcinoma as an ultra-hot subtype and revealing that massive T cell influx in highly infiltrated niches is paradoxically driven into terminal exhaustion, contrasting with severe immune exclusion in cold tumors. Furthermore, stromal-vascular adaptation couples pro-angiogenic tip-endothelial activation with a dynamic opposition between structural extracellular matrix (ECM) deposition and focal proteolytic degradation. Anchored by the pan-cancer MES adaptation, we leveraged subcellular spatial transcriptomics to redefine the microenvironment around four MES-like architectural niches, mapping an evolutionary trajectory from perivascular entry to immunosuppressive stromal remodeling. Finally, projecting these architectures onto independent clinical cohort establishes the MES-S2 proliferative niche as the primary driver of severe clinical deterioration across both transcriptomic and proteomic dimensions.

Cancer Biology↗

bioRxiv2026

Discovery and Development of First-in-Class Cereblon-Recruiting RIPK1 Degraders

Receptor-interacting protein kinase 1 (RIPK1) is a critical regulator of programmed cell death and is implicated in various pathological conditions, particularly in mediating tumor resistance to immune checkpoint inhibitors (ICBs). In this study, we have pioneered the development of a novel cereblon (CRBN)-recruiting RIPK1 degrader, LD5095, through systematic optimization of linker and CRBN ligand portion. LD5095 demonstrates potent and selective RIPK1 degradation across cell lines, with rapid kinetics and sustained degradation over 72h post-washout. Functionally, RIPK1 degradation by LD5095 significantly sensitized Jurkat cells to TNF-induced apoptosis. Furthermore, LD5095 exhibited favorable pharmacokinetics, including metabolic stability and an extended half-life. Strikingly, in vivo, a single dose of LD5095 achieved durable RIPK1 degradation in xenograft tumors over 6 days. These findings underscore the potential of LD5095 as a chemical probe for studying RIPK1 biology and a promising candidate for cancer treatment.

pharmacology and toxicology↗

bioRxiv2026

Isoform-resolved spatial transcriptomics on a lab-made high-density array via a single-chip NGS-TGS workflow

High-resolution spatial transcriptomics is still cost-prohibitive and dominated by short-read sequencing, limiting in situ detection of transcript structures. Here we present a low-cost, lab-made, high-density bead-in-microwell chip assembled by routine centrifuge and decoded by a tri-part combinatorial long-barcode strategy. The design provides 5.6x107 barcode combinations across 4.1x10{square} capture sites, reducing collisions while remaining tolerant to third-generation sequencing raw errors. Using a single-chip, dual-platform workflow, we split full-length cDNA for next-generation and third-generation sequencing within the same spatial coordinates. In an incompatible tomato-pepper graft, we identified interface-associated splicing reprogramming, including thousands of unannotated isoforms, enriched novel CaGRP1 isoforms, and SlPIP2 intron retention, supported by individual long-read sequences. In mouse embryos, we further demonstrate cross-species applicability by resolving unannotated Col1a2 isoforms in oligodendrocyte progenitors and spatially restricted Dalrd3 intron retention. This cost-efficient system broadens access to spatial full-length transcriptomics and may support diverse biological studies.

plant biology↗

bioRxiv2026

PocketGNN: A Cross-Modal Framework Unifying Local 3D Pocket Geometry and Global Sequence Semantics for Enzyme Kinetic Prediction

The enzyme turnover number (kcat) is a pivotal kinetic parameter for understanding bio-catalytic efficiency, yet its accurate prediction remains a grand challenge due to the complex interplay between local physicochemical constraints and global evolutionary context. Existing methods typically bifurcate into sequence-based approaches, which capture evolutionary semantics but miss fine-grained spatial details, or structure-based models, which often suffer from noise in whole-protein representations or lack global context. To bridge this gap, we propose PocketGNN, a cross-modal deep learning framework that synergizes the precision of local 3D geometry with the breadth of global 1D sequence semantics. PocketGNN introduces a high-fidelity graph representation of the active pocket, enriched with a novel 24-dimensional geometric edge encoding (RBF distances, bond angles, dihedral angles) to capture the stereochemical determinants of catalysis. Crucially, this local structural view is fused with global evolutionary information extracted from pre-trained protein language models (ESM-2), creating a unified representation that spans spatial scales. Evaluated on a rigorous dataset derived from IntEnzyDB, PocketGNN achieves a Pearson correlation coefficient (r) of 0.98 and a coefficient of determination (R2) of 0.918 for log10(kcat) under standard random splitting. Furthermore, under a strict 40% sequence identity split designed to test zero-shot generalization to unseen families, the model maintains a robust correlation (r = 0.67, R2 = 0.44), significantly outperforming recent state-of-the-art methods including CatPred (r = 0.52) and CataPro (r = 0.50). Interpretability analysis confirms that the model successfully attends to key catalytic residues, validating its ability to learn chemically meaningful structure-function relationships rather than mere sequence memorization.

bioinformatics↗

bioRxiv2026

Designing AAV Capsid Protein with viability-guided Diffusion Model

Adeno-associated virus (AAV) capsids have shown great promise as delivery vectors for gene therapy. However, the natural properties of AAV capsids impose significant limitations on their efficacy. While capsid library screening and directed evolution have facilitated the development of improved vectors, current methods for generating AAV libraries frequently yield nonviable variants that fail to assemble or package DNA. Here, we propose a viability-guided diffusion model AAVDiffusion for de novo viable AAV capsid design. AAVDiffusion iteratively denoises Gaussian vectors into vectors of AAV capsid protein sequences, yielding intermediate latent variables. By leveraging the continuous nature of these latent variables, AAVDiffusion integrates an additional viability classifier that applies gradient updates to enhance the generation of viable AAV sequences. Through extensive computational tests, AAVDiffusion exhibits superior performance in generating viable AAV sequences. Furthermore, 196 AAV candidates were identified through a selection workflow, with the potential to cross the blood-brain barrier, thereby offering safer and more effective gene therapies for brain diseases. AAVDiffusion offers a powerful and efficient computational method for designing viable AAV capsids, advancing the development of AAV vectors for gene therapy.

bioinformatics↗

bioRxiv2026

An additional site mutation in MITA/STING gain-of-function mutants abolishes the autoimmune SAVI phenotypes and directs a therapeutic strategy.

STING-associated vasculopathy with onset in infancy (SAVI) is an autoimmune disease caused by gain-of-function mutations (GOFs) of MITA/STING and the most frequent GOFs for SAVI are MITAN154S and MITAV155M. However, how MITA GOFs are spontaneously activated remains incompletely understood. Here, we show that the activity of MITA hinge-region GOFs is compromised by an additional mutation at Lys150 and that the SAVI phenotypes of MITAN153S/WT mice are completely abolished in the MITAK150N/N153S (MITANS/NS) mice. Mechanistically, MITA GOFs constitutively associate with iRhom2 for the spontaneous ER-to-Golgi translocation, which is substantially inhibited by the introduction of a mutation at Lys150. Interestingly, cGAMP binds to MITANS, triggers the ER-to-Golgi translocation of MITANS as well as the MITANS-iRhom2 interaction, and induces the expression of downstream genes in MitaNS/NS cells similarly as in Mita+/+ cells. Consistently, structural studies demonstrate an inactive open conformation of apo-MITANS characterized by connector region crossover and a curved filament of cGAMP-bound MITANS characterized by parallel connector regions, similar to those observed in wild-type MITA. Furthermore, we design a SAVI-inhibitory peptide (SIP) that selectively inhibits the interaction between MITAN153S and iRhom2 and the activity of MITA GOFs and thereby abolishes the SAVI phenotypes of the MITAN153S/WT[->]WT chimeric mice. These findings reveal a previously uncharacterized mechanism for the spontaneous activation of MITA GOFs and highlight a potential therapeutic intervention for SAVI.

immunology↗

bioRxiv2025

A Comprehensive Treatment-Induced Resistance Atlas of Glioblastoma Reveals a Fibrotic Niche Shielding the Tumor from Immunotherapy

Therapeutic resistance in IDH-wildtype glioblastoma (GBM) is driven by an intricate interplay between cellular plasticity and protective microenvironmental niches. By constructing the GRIT-Atlas--a massive transcriptomic compendium encompassing nearly one million high-quality single cells across 17 cohorts--and cross-analyzing it with 48 independent Visium spatial transcriptomics sections, we discover a recurrent "Spatial Resistance Triad" comprising mesenchymal-like (MES-like) malignant cells, myeloid-derived suppressor cells (MDSCs), and collagen-secreting cancer-associated fibroblasts (CAFs). This large-scale spatial integration reveals that the triad physically fortifies microvascular proliferation (MVP) and pseudopalisading necrosis (PAN) niches. To substantiate these findings at true single-cell resolution, we deploy high-plex spatial molecular imaging (CosMx SMI) to map over 400,000 single cells across a clinical validation cohort. This high-resolution architecture firmly validates the triads geography and, critically, provides structural gradient evidence unmasking a spatiotemporal continuum wherein PAN emerges as a direct functional consequence of MVP advancement and subsequent vascular collapse. Mechanistically, stromal CAF-derived COL6A1 engages CD44 receptors on MES-like cells to accelerate a potent neural stemness cascade within these protective domains. Utilizing a blood-brain barrier (BBB)-penetrant multi-library compound screen, we identify Lacidipine as a multi-modal stromal disruptor that successfully silences CAF activation and halts matrix secretion. Incorporating Lacidipine into the standard chemo-immunotherapy backbone (TMZ + anti-CSF1R) completely dismantles the protective desmoplastic matrix niche, forcing a profound collapse of intracranial tumor burden and dramatically extending overall survival in orthotopic models. Collectively, our study leverages unprecedented single-cell and spatial scale to provide a definitive blueprint of GBM therapeutic evasion, establishing matrix-targeted intervention as a mandatory prerequisite for successful glioblastoma eradication.

cancer biology↗

bioRxiv2025

MicroGenomer: A Foundation Model for Transferable Microbial Genome Representations Enabling Multi-scale Genomic Understanding and Ecophysiological Trait Prediction

Microorganisms underpin global biogeochemical cycles and represent a vast, underexplored reservoir for sustainable biotechnology and human health. Their diversity, arising from the interplay of numerous genes within holistic genomic contexts, dictates complex ecophysiological traits across varied environments. To bridge the gap between complex genomic sequences and associated biological functions and phenotypic outcomes, we present MicroGenomer, a foundation model for transferable microbial genome representations enabling multi-scale genomic understanding and ecophysiological trait prediction. MicroGenomer leverages a hierarchical training strategy comprising pre-training on large-scale genomic sequences (234.5 billion base pairs), domain-specific mid-training using the GTDB-curated marker gene set, and task-specific post-training. Despite a streamlined architecture of 470 million parameters, MicroGenomer achieves performance on key tasks competitive with models nearly 85 times its size. Advancing beyond established gene-scale encoders, MicroGenomer generates robust embeddings at the genome scale for downstream modeling. Extensive evaluations demonstrate that MicroGenomer effectively captures phylogenetic structures in species space, excelling in geneand genome-scale understanding as well as ecophysiological trait prediction. The practical utility of these capabilities is further demonstrated by targeted wet-lab validation on newly isolated strains, indicating that MicroGenomers predictions provide reliable guidance for biological experiments. Collectively, MicroGenomer offers a high-performance, resource-efficient framework that transforms raw sequence data into actionable biological insights, providing a powerful foundation for microbiome research and biotechnology.

bioinformatics↗

bioRxiv2025

Decoding the fibroblast/mast cell signaling pathway of acupuncture

Acupuncture has been practiced for thousands of years with documented therapeutic effects, yet its underlying biological mechanisms remain poorly understood. Here, we show that acupuncture stimulation converts local mechanical forces into neuronal signals through defined cellular interactions, establishing a causal axis that links micro-level events at the ST36 acupoint to systemic therapeutic effects. We demonstrate that acupuncture needle manipulation transmits mechanical tension through collagen fibers to local fibroblasts, which subsequently secrete stem cell factor (SCF) and interleukin-33 (IL-33) in the acupoint. SCF recruits mast cells, while IL-33 induces their activation and the release of neuromodulatory molecules. These molecules then engage the nervous system, as evidenced by c-fos expression in the lumbar dorsal horn, demonstrating the transmission of acupuncture-induced peripheral signals into central circuits. Together, these findings define a functional axis linking needle-evoked stimulation to neuronal activation and provide a mechanistic framework for understanding the biological basis of acupuncture and for its evidence-based refinement.

cell biology↗

bioRxiv2025

Autolamellasomes: An Autophagy-Dependent Parallel Pathway for Bulk ER Degradation and Lysosome Biogenesis

Lysosomes containing multilamellar membrane whorls are a hallmark of cellular aging and storage disorders, yet the biogenesis of these structures has remained elusive for decades. Here, we identify a distinct form of endoplasmic reticulum (ER) remodeling, termed autolamellasomes, which mediates bulk ER degradation under chronic mTOR inhibition. Unlike canonical ER-phagy, autolamellasomes are concentric ER stacks that form via an autophagy-dependent but receptor-independent mechanism. Using Cryo-ET, CLEM, and a reconstituted cell-free system, we demonstrate that these structures arise from the cytosolic compaction of fragmented ER membranes, driven by the core autophagy machinery. We find that autolamellasomes accumulate in senescent cells and fibroblasts from patients with Hutchinson-Gilford progeria syndrome, linking sustained mTOR suppression to lysosomal membrane homeostasis. Our results resolve the origin of intralysosomal whorls and define a conserved pathway that couples nutrient sensing to membrane turnover and cellular aging.

cell biology↗

bioRxiv2025

Antibody feedback establishes an affinity brake in the germinal center

Recent advances in mRNA vaccine technology have opened the door to novel types of antigen display, but little is yet known as to how B cells recognize and respond to these formats. By delivering an mRNA-LNP encoded membrane-bound immunogen displaying three conserved HIV-1 Envelope (Env) epitopes to knock-in mouse models with B cell receptors (BCRs) of defined affinities, we investigated how epitope-specific competition shapes germinal center (GC) responses. Co-activation of B cells targeting different epitopes did not alter GC kinetics observed in individual activations, but a striking inverse correlation was observed between BCR affinity and GC residence time in either scenario: high-affinity B cells exhibited shorter persistence in GCs, while those with lower affinity to the antigen were maintained. Furthermore, B cells were able to engage in GC reactions at equivalent rates in the presence or absence of clonal lineages binding the same epitope with similar affinities, while higher-affinity clones suppressed lower-affinity counterparts targeting the same epitope. Spatial transcriptomics revealed plasma-like cells within and adjacent to the GC which, together with the detection of early IgG in draining lymph nodes, suggests that local antibody production from these cells may contribute to feedback-driven kinetics. These findings indicate that a self-modulated local antibody feedback loop may act as a "brake" on epitope-specific recognition--dampening further affinity enhancement for high-affinity B cells and facilitating epitope spreading by redirecting the response toward alternative epitopes.

immunology↗

bioRxiv2025

EHMT2 Controls Neural Crest-Derived Craniofacial Development but is Dispensable in Limb Development

Post-translational modifications of histones, such as methylation of histone H3 at lysine 9 (H3K9), play critical roles in regulating chromatin structure and gene expression. EHMT2 (also called G9A), a histone methyltransferase, mediates H3K9 mono- and dimethylation and has been implicated in both transcriptional repression and context-specific gene activation. Although global knockout of the mouse Ehmt2 gene results in early embryonic lethality, tissue-specific knockouts have uncovered diverse roles in organ development. However, how EHMT2 contributes to skeletal development in a lineage-specific manner remains to be fully elucidated. Here, we investigated the role of EHMT2 in skeletal development by conditionally inactivating Ehmt2 in neural crest-derived and mesoderm-derived progenitors using Wnt1-Cre and Prx1-Cre mouse lines, respectively. Loss of Ehmt2 function in neural crest cells led to postnatal growth failure and craniofacial defects, including delayed intramembranous ossification and malformations of the jaw and cranial base. Transcriptomic analysis of neural crest cells revealed disrupted chromatin regulatory networks, reduced expression of proliferation-associated genes, and upregulation of inflammatory pathways. In contrast, inactivation of Ehmt2 in Prx1-expressing mesodermal progenitors had minimal impact on limb and cranial bone development, with no significant alterations in bone mass or osteoblast function. Together, these results reveal a lineage-specific requirement for EHMT2 in neural crest-derived skeletal tissues, suggesting that distinct progenitor populations exhibit differential dependency for bone development.

cell biology↗