Biology subjects

Liu, R.

Publications and source records attributed to Liu, R..

At least 55 records · Page 3Linked to original sources

bioRxiv2026

Salmonella exploits metal-responsive UMPylation to enable intracellular biphasic growth and antibiotic tolerance

Salmonella develops profound antibiotic tolerance during in vivo infection, directly contributing to antimicrobial treatment failure. Although phenotypic heterogeneity in bacterial growth rates underlies this tolerance, the exact mechanisms remain elusive. Here, we reveal that Salmonella exhibits a biphasic growth pattern within macrophages, with initial antibiotic levels modulating lag phase duration. Crucially, this growth plasticity is orchestrated by metal ion homeostasis and a reversible PTM-mediated protein phase transitions. Mechanistically, under stress conditions, enhanced Mn2+ uptake activates the UMPylator YdiU, which modifies critical translational machinery components such as elongation factor TufA. These modifications induce protein aggregation, leading to translational arrest and growth stasis. Conversely, upon stress alleviation, Salmonella accumulates Mg2+ and effluxes Mn2+, then the de-UMPylase YdiV is activated, which resolves protein aggregates to enable rapid bacterial resuscitation and proliferation. This metal-responsive proteostatic switch represents an evolutionarily conserved strategy for bacterial stress adaptation, with broad implications for combating antimicrobial tolerance. SIGNIFICANCE STATEMENTAntibiotic tolerance, a key cause of treatment failure in chronic infections, is often linked to bacterial phenotypic heterogeneity, yet the underlying regulatory mechanisms remain largely understood. This study uncovers a fundamental survival strategy employed by intracellular pathogen Salmonella, where host-derived manganese acts as a critical signal to induce a reversible, dormant state. We demonstrate that stress-induced Mn2+ influx activates widespread protein UMPylation, which functions as a molecular glue to drive the aggregation of essential proteins, thereby halting bacterial growth and conferring antibiotic tolerance. Crucially, this state is reversible upon stress relief through Mn2+ efflux and subsequent de-UMPylation, enabling rapid bacterial resuscitation. As an evolutionarily conserved mechanism of adaptive tolerance, this metal-ion-responsive switch, mediated by reversible protein UMPylation, identifies a key target for developing therapeutics against persistent bacterial infections.

microbiology↗

bioRxiv2026

Near completely reversing the γ- to β-globin switch by enhancer release, retargeting and reinforcing

The {gamma}- to {beta}-globin switch is intricately regulated during human ontogeny, and this process is manipulated for therapeutic approaches to treat {beta}-hemoglobinopathies by activating {gamma}-globin expression. Several genetic strategies to reactivate HbF have partially reversed the {gamma}- to {beta}-globin switch and ameliorated the clinical symptoms of {beta}-hemoglobinopathies. However, whether the {gamma}- to {beta}-globin switch can be completely reversed remains unknown. Completely reversing the {gamma}- to {beta}-globin switch requires a thorough redirection of the locus control region (LCR) from interacting with the {beta}-globin gene (HBB) to interacting with the {gamma}-globin gene (HBG). Here, we found that disrupting the KLF1-mediated HBB-LCR interaction by mutating the CACCC motif in HBB leads to the release of the LCR and its retargeting to other {beta}-like globin genes. Moreover, simultaneously disrupting the KLF1-mediated HBB-LCR interaction and the epigenetic repression of HBG by combined editing of the CACCC motif in HBB and the TGACCA motif in HBG reinforces the HBG-LCR interaction, resulting in almost exclusive {gamma}-globin expression while nearly absent {beta}-globin expression, achieving near complete reversal of the {gamma}- to {beta}-globin switch. This finding demonstrates the comprehensive regulation of the {gamma}- to {beta}-globin switch by gene competition and gene silencing mechanisms. This finding also suggests that silenced genes can be fully activated through the redirection of enhancer-promoter contacts and that the specificity of enhancer-promoter contact within chromosomal domains is achieved through the transcription factor clusters binding to enhancers and promoters. Combined editing of the CACCC&TGACCA motifs also offer a more optimal therapeutic strategy for {beta}-hemoglobinopathies.

molecular biology↗

bioRxiv2026

Molecular determinants of antibody-mediated priming to enhance detection of ctDNA

Liquid biopsies can enable cancer detection and monitoring yet remain limited by low concentrations of ctDNA. To address this limitation, we previously introduced a monoclonal antibody (mAb) "priming agent" that transiently increased the concentration of ctDNA in blood. Here, we investigated the molecular features that drive this effect. In a panel of novel mAbs that bound cfDNA, both those targeting dsDNA and those targeting mononucleosomes increased the concentration of ctDNA. mAbs with high avidity to dsDNA performed best, suggesting dsDNA as the key binding target. One agent preserved short, biologically informative molecules of cfDNA typically depleted at baseline. The Fc domain was dispensable, as F(ab)2 fragments retained priming activity. Leveraging these insights, we engineered single-chain agents using the dsDNA binding domain sso7d, expanding priming strategies beyond immunoglobulins. This study identifies the molecular features and general design principles for mAb-based priming agents to enhance recovery and detection of ctDNA. SignificanceLow concentration of ctDNA limits the sensitivity of liquid biopsies in many applications. This study shows how engineered antibodies and other dsDNA-binding molecules can inhibit clearance of ctDNA from the bloodstream and increase concentration of ctDNA in a blood draw, and identifies the key features that enable this activity. Modulation of cfDNA in the bloodstream using these agents can increase recovery of ctDNA and improve the performance of liquid biopsies for cancer detection.

bioengineering↗

bioRxiv2026

A Modularized Stretchable Microneedle Array for Minimally Invasive and Efficient Drug Delivery

A novel Modularized Stretchable Microneedle Array (MSMA) is developed for minimally invasive and efficient drug delivery. By applying a modular design, the MSMA can be highly customized, offering versatility for different therapeutic applications. Its high stretchability allows for simultaneous stretching of both the microneedle and the skin, enhancing the penetration rate. Evaluations using tissue-mimicking materials and porcine skin demonstrated significant improvements in penetration efficiency and reduced residual strain. The MSMA achieved penetration rates of up to 89% on porcine skin under a minimum push-down distance of 2 mm. Moreover, the MSMAs ability to stretch and release in synchronization with the skin significantly reduces residual strain. This innovative platform promises reduced pain and discomfort, representing a substantial advancement in transdermal drug delivery technology.

bioengineering↗

bioRxiv2026

Synthetic Hybrid Receptors for Safer and Programmable T Cell Therapy

Engineered T cell therapies have achieved significant clinical success in hematological malignancies but remain largely ineffective in solid tumors. Overcoming this limitation requires strategies that enhance T cell function while avoiding systemic immune toxicities and pathological T cell states. Existing approaches typically rely on constitutive gene overexpression or suppression to augment potency or remodel the tumor microenvironment, but these strategies frequently lead to dysregulated immune activation and dose-limiting toxicity. Here, we present Hybrid Receptors (Hybrid-Rs), a modular receptor platform that integrates features of chimeric antigen receptors (CARs) and SyNthetic Intramembrane Proteolysis Receptors (SNIPRs) to couple antigen-dependent T cell activation with programmable gene regulation. Hybrid-Rs enable precise, context-dependent control of T cell potency, differentiation states, and conditional expression of secreted immunotherapeutic payloads with otherwise prohibitive toxicity. Hybrid-Rs are readily humanized and compatible with precision genome editing in primary human T cells, providing a direct and practical path to clinical translation.

bioengineering↗

bioRxiv2026

CTLA-4 inhibitors drive colitis through metabolic reprogramming-mediated Treg/Th17 imbalance

Immune checkpoint inhibitors (ICIs), especially CTLA-4 inhibitors (CTLA-4), exhibit a high incidence of colitis as an immune-related adverse event (irAE) during cancer treatment, severely limiting patient benefit. Clinically, both treatment interruption and existing intervention drugs for ICI-mediated colitis may compromise antitumor efficacy. However, there is inadequate research on the pathogenesis of ICI-mediated colitis, with findings often conflicting. Here, we first established multiple clinically relevant animal models, including an immuno-humanized ICI-mediated colitis model. Through time-series transcriptomics, we discovered that CTLA-4-induced colonic toxicity exhibits characteristics ranging from early metabolic reprogramming represented by glycolysis to later immune disorders represented by Th17 responses. By targeting colonic CTLA-4+ T cells, CTLA-4 blocked CD80/CD86-CTLA-4 interaction, thereby activating the PI3K-AKT-mTOR pathway. Subsequently, mTOR mediated metabolic reprogramming in T cells, shifting them from Treg-biased oxidative phosphorylation to Th17-biased glycolysis. The colonic toxicity of CTLA-4 has also been demonstrated to depend on the PI3K-AKT-mTOR pathway, glycolysis, and Th17 responses. Notably, metformin significantly relieved ICI-mediated colitis by inhibiting mTOR without impeding antitumor efficacy. Collectively, these findings highlighted the metabolic-immune axis in the colonic toxicity of ICI and provided a clinically superior intervention strategy.

pharmacology and toxicology↗

bioRxiv2026

A novel anaerobic n-Hexadecane hydroxylation pathway in Thermalkanevorax longiformis gen. nov., sp. nov. isolated from a deep-sea hydrothermal vent

Hydrothermal systems contain a wide spectrum of alkanes derived from biological sources and geological processes. However, despite their widespread occurrence in hydrothermal environments, the capacity and mechanisms of anaerobic alkane degradation by members of the phylum Bacillota remain poorly understood. In this study, we isolated a bacterial strain (designated strain L01) from the Lost City hydrothermal field (LCHF) by adding crude oil for enrichment culture. Strain L01 can grow anaerobically at 55 {degrees}C and pH 7.0 in medium supplemented with n-hexadecane, indicating its capability for anaerobic alkane degradation. Phylogenetic analysis based on the 16S rRNA gene sequence (93.11% similarity to its closest described relative) and average amino acid identity (AAI, 66.27-66.58%) revealed that strain L01 represents a novel genus within the family Symbiobacteriaceae, for which the name Thermalkanevorax longiformis gen. nov., sp. nov. is proposed. Cells of strain L01 are elongated rods measuring approximately 10-20 m in length and [~]0.2 m in diameter. Integrated genomic, transcriptomic, and metabolomic analyses revealed a previously unrecognized anaerobic n-hexadecane utilization pathway, involving initial alkane activation by AhyA homologous enzymes followed by stepwise oxidation and {beta}-oxidation. Together, these findings provide new insights into anaerobic alkane metabolism by hydrothermal bacteria, expanding current understanding of microbial hydrocarbon degradation in extreme marine environments. IMPORTANCEAlthough alkanes have be reported degradation by microorganisms through both aerobic and anaerobic metabolic pathways in hydrothermal ecosystems, the mechanisms underlying anaerobic alkane degradation remain incompletely understood, and the metabolic strategies employed by hydrothermal microorganisms are still poorly characterized. In this study, a novel genus bacterium Thermalkanevorax longiformis within the Symbiobacteriaceae family was isolated from the LCHF and identified an unrecognized C1-position n-hexadecane hydroxylation of AhyA homologous proteins at the anaerobic condition. These findings not only provide fundamental insights into the metabolic strategies of hydrothermal bacteria but also expand our understanding of microbial hydrocarbon degradation in extreme marine environments, offering a potential genomic resource for biotechnological applications in hydrocarbon remediation.

microbiology↗

bioRxiv2026

Time-dependent Glucocorticoid-Induced Transcriptomic Changes in Human Trabecular Meshwork and Schlemm's Canal

PurposeTo identify the transcriptomic changes induced by dexamethasone (DEX) in trabecular meshwork (TM) and Schlemms canal endothelial (SCE) cells with RNA-sequencing (RNA-seq). MethodsHuman TM (n=10) and SCE cell strains (n=5) were isolated from healthy donor eyes and exposed to DEX 100nM and vehicle (control). Three DEX exposure times were evaluated: 1-hour, 6-hours, and 2 days. RNA-seq was performed on Illuminas TruSeq platform and gene expression was quantified using featureCount. DESeq2 paired (treated and untreated) sample test was applied to identify genes transcriptionally responsive to DEX (DEGs) at false discovery rate <0.05. Gene-set enrichment analyses were performed on DEGs. DEGs were tested for association with glaucoma (POAG) and intraocular pressure (IOP). ResultsNine TM and 4 SCE strains passed quality control. After 2-day DEX exposure, there were 857 and 2,086 DEGs in TM and SCE, respectively. Of these, 411 genes were differentially expressed in both TM and SCE, including FKBP5 (17.3-fold-change, p=6.9x10-53) and FAM107A (25.1-fold-change, p=4.0x10-240), the most significant DEG after 2-day DEX exposure in TM and SCE, respectively. The 2-day DEX DEGs in TM and SCE were enriched in cell adhesion, extracellular matrix, and response to stimulus in Gene Ontologies (p<3.7x10-6). Early response DEGs were enriched in immune-related processes. Thirteen DEGs in TM were significant at all three time points, including PER1. LTBP2 is a TM-only DEG and FAM105A a SCE-only DEG associated with IOP and POAG risk. ConclusionsThis study identified candidate genes and pathways for glucocorticoid-induced ocular hypertension which can be further explored in human genetic analyses.

genomics↗

bioRxiv2025

Taxane-Induced Conformational Changes in the Microtubule Lattice Activate GEF-H1-Dependent RhoA Signaling

Taxanes are widely used chemotherapeutic agents that perturb cell division. They also exert effects during interphase, but the underlying mechanisms are poorly understood. Here, we show that taxanes activate RhoA signaling and induce actin remodeling by displacing the RhoA activator GEF-H1 from microtubules. This taxane-induced release of GEF-H1 occurs rapidly, is independent of tubulin post-translational modifications, and can be recapitulated using purified proteins. In vitro reconstitution assays combined with analyses of microtubule structure revealed that microtubule binding by GEF-H1 is inhibited by microtubule-stabilizing agents that expand the microtubule lattice, such as taxanes and GMPCPP, but not by others, including GTP{gamma}S and discodermolide, which stabilize a compacted microtubule lattice. Our findings demonstrate that alterations in microtubule lattice conformation can activate key signaling pathways, offering new insights into the mode of action of taxanes and the possible origins of their side effects.

cell biology↗

bioRxiv2025

Benchmarking computational methods for multi-omics biomarker discovery in cancer

Multi-omics profiling characterizes cancer biology and supports biomarker discovery for prognosis and therapy selection. Although numerous computational multi-omics biomarker identification methods have been proposed, their ability to identify clinically relevant biomarkers has not been systematically evaluated, leaving it unclear whether the resulting biomarker nominations are reliable for downstream validation. Here we systematically benchmark 20 representative statistical, machine learning and deep learning methods using curated gold-standard prognostic and therapeutic biomarkers across five real-world datasets. We evaluate performance in terms of both biomarker identification accuracy and stability. Overall, DeePathNet and Deep-KEGG achieve the best performance. Across methods, effective biomarker recovery is associated with the integration of biological knowledge, global feature interactions, multivariate feature attribution, and effective regularization. Analysis of omics type contributions reveals method- and modality-specific biases, highlighting the importance of broader omics integration. We further evaluate methods on simulated datasets to probe sensitivity with controlled signal and noise. By aggregating results from top-performing methods, we construct consensus biomarker panels that nominate candidates for potential investigations. Finally, we provide user-friendly interfaces to allow researchers to benchmark new methods against the 20 baselines or apply selected methods for biomarker identification on custom multi-omics datasets. Our benchmark is publicly available at https://github.com/athanzli/CancerMOBI-Bench.

bioinformatics↗

bioRxiv2025

A spatially explicit delineation of transition zones for biogeographic regionalization

O_LIA central goal of biogeography is to partition the worlds biota into meaningful biogeographical regions. While naturalists have long delineated biogeographic regions using hierarchical hard clustering approaches, they have often acknowledged the existence of transition zones. However, such transition zones have not been empirically identified or distinguished from hard clusters. Transition zones are outstanding areas for understanding the processes that shape biotic assembly; however, a coherent framework for delineating them is lacking. C_LIO_LIHere, we propose a framework to empirically identify and analyze transition zones through a new metric, the "Transition Zone Index (TZI)". The TZI decomposes the proportional contribution of multiple biotas to a spatial unit (called a map cell) derived from a Grade of Membership model into a single quantitative measure per cell using Shannon entropy principles. The transition zones can then be quantified and presented as a gradient, scaled from TZI = 0 (indicating distinctive bioregions) to TZI = 1 (corresponding to the highly transitional areas with maximally admixed biotas). Finally, we implement spatial regression models to detect the predictors of these transition zones. C_LIO_LIWe demonstrate the application of our framework using vascular plants of southern Africa. Through the spatial variation of TZI, our approach represents transition zones as a continuous gradient of biotic turnover between bioregions. These transition zones were found to be either previously delineated as discrete clusters or incorporated into neighboring clusters. Our approach effectively captures both the intensity and width of transitional dynamics, showing that these patterns are primarily shaped by precipitation. C_LIO_LIBiogeographic transition zones are a measurable gradient, and our framework provides a robust quantitative foundation for detecting and analyzing the drivers of these areas. This improves our ability to understand the patterns and processes underlying biogeographic transition and provides important implications for conservation planning and biodiversity management. C_LI

ecology↗

bioRxiv2025

Benchmarking of duplex sequencing approaches to reveal somatic mutation landscapes

Detecting somatic mutations in normal tissues is challenging due to sequencing errors and the low allele fractions of post-zygotic variants. Duplex sequencing greatly reduces errors and can detect mutations at any allele fraction, but systematic, cross-platform comparisons are lacking. We present a comprehensive benchmarking of six duplex sequencing technologies used by the SMaHT Network: CODEC, CompDuplex-seq, HiDEF-seq, NanoSeq, ppmSeq, and VISTA-seq. We evaluated their performance using cord blood DNA, a tumor-normal cell line mixture, and homogenates from six human tissues. Each method shows distinct profiles in genomic footprint, sensitivity, and cost. Despite differences in library construction and sequencing platforms, estimates of mutation rates and mutational signatures are highly concordant. Integration with ultra-deep whole-genome sequencing shows that duplex approaches sensitively capture mutations and signatures beyond embryonic or clonally expanded variants. These results provide a foundation for selecting duplex methods and interpreting their data, enabling scalable single-molecule analyses of somatic mutation landscapes. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=101 SRC="FIGDIR/small/692823v1_ufig1.gif" ALT="Figure 1"> View larger version (46K): org.highwire.dtl.DTLVardef@123b221org.highwire.dtl.DTLVardef@83ba0aorg.highwire.dtl.DTLVardef@2b0ab0org.highwire.dtl.DTLVardef@1cae35e_HPS_FORMAT_FIGEXP M_FIG C_FIG

genomics↗

bioRxiv2025

In Situ Photoactivated Hydrogel Adhesive Dressings for Post Colon Polypectomies (PolypCures)

Colon polypectomy is a widely performed endoscopic procedure that reduces the incidence of colorectal cancer but leaves exposed colonic wounds susceptible to bleeding, perforation, and infection. The current standard of wound care, mechanical clips, is limited by technical complexity, high cost, and poor efficacy for large (> 2 cm in diameter) or difficult-to-access lesions. Here, we present PolypCure, an in situ photoactivated hydrogel adhesive dressing delivered via a single catheter, through either oozing or spraying methods, to achieve rapid, spatiotemporally controlled sealing and hemostasis (within 2 min) of colon lesions using the white light source integrated into standard endoscopic instruments. The hydrogel, composed of norbornene- and thiol-functionalized polyethylene glycol, carboxymethyl cellulose, and Eosin Y, exhibits mechanical properties comparable to colon tissue (G [~] 6 kPa), strong shear strength (> 15 kPa), low swelling (< 200%), and small pore size ([~] 20 {micro}m), ensuring long-term stability (up to 30 days) and protection against bacterial infiltration. The formulation also demonstrates biocompatibility and hemocompatibility. In an in vivo pig colon endoscopic mucosal resection (EMR) model, PolypCure fully adheres to large lesions for at least 3 days and promotes early-stage wound healing. Overall, PolypCure is a promising solution to wound management post-polypectomy.

bioengineering↗

bioRxiv2025

Dorsomedial hypothalamus cold-sensitive neurons drive shivering thermogenesis

Hypothermia, a decline in core body temperature below 35{degrees}C in homeotherms, elicits rapid, asynchronous involuntary shivering thermogenesis; however, the hypothalamic neurons responsible for initiating this response remain uncharacterized. Here, retrograde transsynaptic tracing combined with RNA-seq identified a distinct population of VGLUT2-positive neurons enriched with Calbindin D28k (CALB1) within the dorsomedial hypothalamic nucleus (DMHCalb1/Vglut2). Convergent evidence from chemogenetic, optogenetic, fiber photometry calcium imaging, and electromyography (EMG) experiments demonstrated that these neurons selectively mediate hypothermia-triggered shivering thermogenesis. Bidirectional perturbation of DMHCalb1/Vglut2 activity modulated shivering intensity, while CALB1-negative VGLUT2-expressing DMH neurons were responsible for non-shivering thermogenesis (NST). Genetic ablation of afferent inputs from LSI/POA/VLPAG projecting to DMHCalb1/Vglut2, as well as ex vivo pharmacological synaptic blockade, failed to eliminate shivering responses during hypothermia, indicating that their activation is independent of canonical cold-responsive circuits. Hypothermia significantly suppressed transient A-type K (IA) currents, without affecting delayed rectifier K (IK) currents, accompanied by a pronounced reduction in the IA inactivation time constant relative to normothermic 36 {degrees}C. Collectively, hypothermia elevates DMHCalb1/Vglut2 excitability through IA-dependent reshaping of action potential dynamics. Shortened action potential repolarization and afterhyperpolarization duration cooperatively boost neuronal firing, constituting an intrinsic mechanism for DMHCalb1/Vglut2 neuron-driven shivering initiation.

neuroscience↗

bioRxiv2025

Robotic Handling Preserves Induced Pluripotent Stem Cell Derived Vascular Smooth Muscle Cells Differentiation using a Weekend-free, Automatable, Low-variability, Low-cost (WALL) Protocol

Vascular disease modeling and tissue engineering require scalable, efficient, and reproducible methods for differentiating cells, however no studies have validated whether automation, essential for scale-up, preserves differentiation quality in cells. Here, we developed a differentiation protocol for induced pluripotent stem cell (iPSC)-derived vascular smooth muscle cells (iVSMCs) that is applicable to both manual, and automated robotic culture. iPSC lines generated from male and female donors across various ages were used to optimize the generation of iVSMCs. iVSMCs had consistent morphology and protein expression was largely similar to primary VSMCs, including expression of myosin heavy chain-11 (MYH11), -smooth muscle actin (-SMA), transgelin (TAGLN) and calponin (CNN1). Functionally, these iVSMCs could respond to the vasoconstrictor carbachol. Coupling this protocol with an automated Hamilton liquid-handling robotics system allowed the generation of iVSMCs in large quantities, morphologically similar to manually differentiated iVSMCs. Comparative proteomic analysis confirmed protein expression did not differ between automated and manual differentiation methods. Automation markedly reduced manual labor and facilitated increased production without sacrificing cell quality. This study demonstrates the feasibility of automating iVSMC differentiation, marking a significant step towards scalable VSMC manufacturing for three-dimensional applications, and for the ever-growing demands of organoid and tissue engineering applications. Statement of significanceThe rapidly expanding fields of tissue engineering and high-throughput disease modeling require large quantities of viable cells, making automation essential. Robotic liquid handling systems are increasingly employed for generating large quantities of induced pluripotent stem cell (iPSC)-derived cells, requiring the development of scalable protocols. No studies have yet evaluated whether the use of robotic liquid handling alters the differentiation of cells compared to traditional manual handling. Here, we report a scalable differentiation protocol that robustly generates iPSC-derived VSMCs (iVSMCs) from a cohort of female and male donors of varying age (n=5 lines). We demonstrate for the first time that iVSMCs, differentiated through an automated robotic pipeline, had no significant differences in morphological or proteomic phenotype compared to manual differentiation. The ability to automate this protocol facilitates robust and reproducible differentiation of large quantities of iVSMC, setting a precedent for the application of automation in generating iPSC-derived cells. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=57 SRC="FIGDIR/small/690512v1_ufig1.gif" ALT="Figure 1000"> View larger version (18K): org.highwire.dtl.DTLVardef@673d61org.highwire.dtl.DTLVardef@1eead89org.highwire.dtl.DTLVardef@c3089eorg.highwire.dtl.DTLVardef@6af2cd_HPS_FORMAT_FIGEXP M_FIG C_FIG

developmental biology↗

bioRxiv2025

Agentic Lab: An Agentic-physical AI system for cell and organoid experimentation and manufacturing

Reproducibility in biological research and manufacturing remains constrained by the complexity of multi-step protocols, fragmented data-analysis pipelines, and the intrinsic variability of experimental execution. Here, we present Agentic Lab, an agentic-physical AI platform that unifies large language model and vision language model (LLMs/VLMs)-driven reasoning with real-world laboratory operations. Agentic Lab uses multi-agent orchestration architecture, comprising of specialized subagents for knowledge retrieval, protocol design, multimodal data analysis, and training-free segmentation and representation learning for intrinsically explainable single-cell and organoid phenotyping. These agents operate under the orchestration of a virtual principal investigator MolAgent that is linked to an augmented reality (AR)-based physical AI interface, which can bridge digital reasoning with human physical execution. Agentic Lab perceives real-world experimental activities, provides context-aware instructions, identifying procedural errors in real time for humans to correct, and continuously evolves with its long-term memory database expanding through the accumulation of experimental data logs from human scientists. This interaction allows scientists and AI agents to collaborate and co-evolve dynamically, closing the loop between planning, action, and analysis in the traditional cell and organoid research lifecycle. We demonstrate Agentic Lab in organoid differentiation from human pluripotent stem cells, where it autonomously generates protocols, monitors culture procedures, and identifies subtle morphological heterogeneity linked to growth conditions. The system interprets these phenotypes, grounds them in literature, and proposes targeted instructions for improving differentiation efficiency. By combining multi-agent reasoning with physical laboratory awareness, Agentic Lab transforms experimentation and biomanufacturing from a static workflow into an adaptive, feedback-driven, bidirectional process that integrates agentic AI into the research lifecycle. This framework establishes a foundation for intelligent laboratories that integrate design, execution, and interpretation within a unified agentic-physical system.

bioengineering↗

bioRxiv2025

HEIMDALL: A Modular Framework for Tokenization in Single-Cell Foundation Models

Foundation models for single-cell RNA-sequencing (scRNA-seq) data are emerging as powerful tools for single-cell analysis, yet their performance depends critically on how cells are tokenized into model inputs. Single-cell data lack a canonical tokenization scheme, and many design choices in current single-cell foundation models (scFMs) remain heuristic, entangled, and difficult to evaluate. Here, we introduce HO_SCPLOWEIMDALLC_SCPLOW, a unified framework for dissecting and redesigning tokenizers in scFMs. By decomposing existing tokenization strategies into individual design choices, HO_SCPLOWEIMDALLC_SCPLOW enables attribution of the components that underlie robust generalization, allowing more principled design of improved tokenizers. Combining HO_SCPLOWEIMDALLC_SCPLOW with a minimal transformer backbone, we find that tokenizer design is instrumental for generalization in challenging distribution-shift settings such as cross-tissue, cross-species, and cross-gene-panel cell type classification, as well as reverse perturbation prediction. We show that, while tokenizer choice has little effect in scenarios with matched train and test data, it becomes imperative under distribution shift. Rather than identifying a single globally optimal tokenizer, HO_SCPLOWEIMDALLC_SCPLOW reveals that robust transfer depends on a small number of tokenization design axes - especially gene identity, expression encoding, and ordering - that expose different biological priors to the model. In this sense, universal transferability in scFMs still depends on a non-universal tokenizer interface. Together, these findings establish tokenization as a critical design axis in scFMs and provide design principles and reusable infrastructure for more robust scFMs.

bioinformatics↗

bioRxiv2025

MEDUSA: Maintaining Entire DNA Duplexes for Utmost Sequencing Accuracy

Highly accurate DNA sequencing is essential for many applications. Duplex sequencing-based methods achieve unparalleled accuracy by requiring reads from both strands of the original DNA duplex to match. Yet, methods to prepare dsDNA for sequencing may resynthesize portions of each DNA duplex and cause base damage errors on one strand to become indistinguishable from true mutations on both strands. Here, we report MEDUSA (Maintaining Entire DNA Duplexes for Utmost Sequencing Accuracy) which minimizes dsDNA resynthesis to maximize duplex sequencing accuracy and yield. MEDUSA carefully repairs and blunts fragmented dsDNA, then employs apyrase to digest residual dNTPs, followed by restricted dA-tailing to prevent resynthesis. MEDUSA affords full genome coverage in a simplified protocol that is broadly compatible with dsDNA fragmentation and library preparation kits. We benchmarked MEDUSA on sheared genomic DNA and cell-free DNA and found a residual SNV frequency within a median 1.23-fold (range 0.92-1.88; p < 0.001) of what was expected if resynthesis was mostly blocked, but with full genome coverage and duplex yields within a median 1.02-fold (range 0.33 - 1.46; p = 0.258) of traditional methods that do not limit resynthesis. In all, MEDUSA could enable high breadth or depth of duplex sequencing while limiting false discovery. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=111 SRC="FIGDIR/small/687202v1_ufig1.gif" ALT="Figure 1"> View larger version (12K): org.highwire.dtl.DTLVardef@1ddbc77org.highwire.dtl.DTLVardef@802ca9org.highwire.dtl.DTLVardef@f443ecorg.highwire.dtl.DTLVardef@974bdb_HPS_FORMAT_FIGEXP M_FIG C_FIG

genomics↗