Biology subjects

Guan, J.

Publications and source records attributed to Guan, J..

At least 19 recordsLinked to original sources

bioRxiv2026

AI-guided discovery of natural pocket proteins for portable small-molecule sensing

Recognition-element development remains an important practical challenge in small-molecule biosensing. Here, we present DWCLASS, an AI-guided workflow that combines screening of natural pocket proteins, molecular dynamics simulations, experimental binding assessment, and portable metasurface plasmonic sensing. Using thiamine (vitamin B1, thiamine) and 25-hydroxyvitamin D3 [25(OH)D3] as model analytes, the workflow identified binding responses in 7 of 16 tested analyte-protein pairs (43.8%). Selected proteins were incorporated into a competitive MetaSPR assay with gold nanoparticle amplification and acoustic mixing. The portable device provided a 240 s measurement step for prepared samples. Serum LODs were 2.98 ppt for thiamine and 3.82 ppt for 25(OH)D. Preliminary measurements of human serum specimens were compared with commercial ELISA to explore performance in biological samples. This study establishes an integrated route from AI-assisted screening of natural proteins to portable assay development for two chemically distinct vitamin analytes. Applicability to additional targets and clinical reference accuracy remain to be established.

bioinformatics↗

bioRxiv2026

Polyamines control inorganic polyphosphate levels during bacterial nitrogen starvation

Inorganic polyphosphate (polyP) is a ubiquitous, multifunctional biomolecule that supports stress survival in bacteria. During nitrogen (N) starvation, Escherichia coli rapidly accumulates polyP, which drives the formation of RNA protein granules that sustain long term survival. What triggers this upregulation has remained unknown. Here we show that polyamines, a group of highly conserved, amino acid-derived polycations, regulate polyP accumulation during N starvation. We found that deleting all nine polyamine biosynthesis genes elevates polyP during exponential growth and by more than 5fold under N starvation, with shorter chains preferentially accumulating. To dissect which polyamine matters, we further deleted the three catabolic enzymes that consume polyamines as a nitrogen source, which enabled complementation assays. Supplying any one of the three physiological polyamines, putrescine, spermidine, or cadaverine, restored wild type polyP levels during N starvation. We found that the absence of polyamines did not affect the steady state levels of either the polyP synthesizing kinase PPK or the exopolyphosphatase PPX, pointing instead to posttranslational control of enzyme activity or to as yet unidentified polyamine dependent regulators. These findings establish a direct metabolic link between two universal, functionally intertwined biomolecules and suggest that the polyamine decline upon N starvation contributes to the observed polyP accumulation.

microbiology↗

bioRxiv2026

The filopodial scaffold polyphosphate dictates cell adhesion-versus-invasion decisions

Inorganic polyphosphate (polyP) is an ancient polymer conserved across all life, serving cell type and location specific functions in every major compartment. Yet its role at the plasma membrane, where it accumulates to peak levels in many primary cells, is largely unknown. Here we identify polyP as a stabilizing component of filopodia, actin based membrane protrusions that govern cell adhesion, contact inhibition, and chemotaxis. Elevating cellular polyP increases filopodial stability and enhances cell adhesion, whereas reducing polyP accelerates filopodial disassembly and promotes cell migration. Mechanistically, we find that polyP acts as a structural filopodial scaffold, recruiting and organizing IRSp53, a membrane curvature inducing protein. We show that metastatic fibroblasts and breast cancer organoids carry markedly reduced and intracellularly reorganized polyP levels relative to their non transformed counterparts. Restoring endogenous polyP via lipid nanoparticle delivery suppresses their invasive phenotypes and reverses prometastatic gene expression signatures, implicating polyP as a primordial tumor suppressor.

cell biology↗

bioRxiv2026

VirProtRAG: Literature-grounded viral protein function annotation with retrieval-augmented generation

Viruses play indispensable roles in ecosystems and human health, yet deciphering their molecular functions remains challenging. Many viral protein annotations are incomplete or poorly characterized. Existing tools typically predict functional categories without linking to verifiable evidence, hindering the credibility of functional interpretation. Here, we present VirProtRAG, a viral protein function annotation framework that integrates information retrieval with evidence-grounded knowledge generation. It introduces three task-adapted components: a hybrid retrieval module combining keyword-based and semantic dense retrieval to maximize literature coverage, synonym-expanded and rank-aware retrieval with reciprocal rank fusion for improved search effectiveness, and literature quality and evidence-oriented re-ranking to enhance reliability and interpretability. Results show that hybrid retrieval strategy performed best, with quality and evidence features further enhancing re-ranking. Compared with direct LLM prompting without retrieved literature, it consistently improves generation performance, underscoring the critical role of external knowledge. Finally, we built a searchable database comprising all 17,484 reviewed Swiss-Prot viral proteins, supporting both sequence- and text-based queries. VirProtRAG introduced 32.53% non-overlapping function annotations beyond existing expert curation, and independently supported 56.34% of sequence-inferred function points with retrieved literature. Case studies further demonstrate its capability to augment and refine the characterization of previously unannotated or poorly understood viral proteins. Key MessagesO_LIVirProtRAG bridges large-scale biomedical literature and viral protein annotation, enabling traceable and biologically interpretable annotations grounded in verifiable publications. C_LIO_LIBy integrating hybrid sparse-dense retrieval, synonym expansion, and adaptive re-ranking using literature quality and evidence-type features, VirProtRAG substantially improves recall and ranking of biologically relevant publications, helping ground LLM outputs in traceable literature evidence. C_LIO_LILeveraging this framework, VirProtRAG automatically annotated all 17,484 reviewed Swiss-Prot viral proteins under the Virus taxonomic group and established an interactive, continuously expandable database. The resource may assist curators in rapid validation while supporting large-scale studies of viral evolution and host interactions. C_LI

bioinformatics↗

bioRxiv2026

GR-SAFS: A Graph-Regularized Stacking Framework with Adaptive Feature Selection for High-Dimensional Prognostic Biomarker Discovery

Identifying prognostic biomarkers from high-dimensional transcriptomic data poses a triple challenge: achieving sparsity, preserving biological network topology, and integrating complementary nonlinear signals. Existing methods typically ignore network structure, miss nonlinear interactions, or lack a principled mechanism to fuse heterogeneous model outputs. We introduce GR-SAFS (Graph-Regularized Stacking with Adaptive Feature Selection), a framework with three modules: a Graph-Lasso engine embedding gene co-expression network Laplacian priors, run in parallel with a Random Forest engine; an empirical cumulative distribution function (eCDF) alignment layer that places sparse and dense importances on a common percentile scale; and a diversity-penalized quadratic programming router whose strict convexity yields a unique global optimum. On the TCGA-LUAD cohort, GR-SAFS identifies a 20-gene signature with a training concordance index of 0.700. Across two independent cross-platform microarray cohorts, GR-SAFS is the only method whose frozen signature retains statistically significant risk stratification in every cohort, where stronger-C-index baselines lose significance on at least one external cohort. Functional enrichment anchors the signature to a coherent Wnt/{beta}-catenin axis. An open-source implementation is released for full reproducibility.

bioinformatics↗

bioRxiv2026

Temporal multi-omic profiling of immune, gut, and microbiome responses to ischemic stroke reveals convergence of host and microbial perturbations one week after brain injury

Ischemic stroke poses a significant medical challenge with limited therapy options, therefore detailed understanding of stroke-associated pathophysiological processes across systems and organs is crucial for further research and identification of novel therapeutic targets. In this manuscript, we provide parallel multi-omic host and gut microbiome characterization on several timepoints (day 1, 7 and 14) in the mouse experimental stroke model (middle cerebral artery occlusion, MCAo). Expanding existing host-derived datasets, we profiled transcriptomes from microglia, brain-infiltrating leukocytes and peripheral leukocytes using single cell RNA sequencing. Our data deliver time-resolved characterization of microglial subtypes and highlight heterogeneous dendritic cell populations as main interaction partners of microglia on all timepoints. In peripheral blood, we did not observe large transcriptomic differences when comparing the immune subsets from MCAo and sham-operated control animals. Here, the neutrophils exhibited most transcriptomic changes on day 1 among all blood leukocytes. Parallel proteomic analysis of 5 intestinal segments (duodenum, jejunum, ileum, caecum and colon) and mesenteric lymph nodes highlighted day 7 as the most important timepoint for changes in the gut-related metabolic pathways especially in the jejunum and colon. Specific hypothesis testing revealed compartmentalized regulation of gut-related immune pathways and proteins related to gut permeability. Finally, gut microbiome analyses (longitudinal metatranscriptomics including day -1, 3, 7, 14, and metagenomics from day 14) highlighted temporally matched changes in microbial gene expression (with day 7 emerging again as the most relevant timepoint), larger overall community perturbations when compared to baseline from day -1 in stroke animals and expansion of facultative anaerobes on day 7. Graphical abstractWe performed longitudinal multi-omic host and gut microbiota profiling from experimental stroke and sham control mice. Our data indicate marked differences in brain-infiltrating leukocytes and microglia already on early timepoints after surgery, whereas stroke-specific changes in gut proteomics, bacterial gene expression and community structure emerge on day 7 after middle cerebral artery occlusion/sham surgery. Created in https://BioRender.com O_FIG O_LINKSMALLFIG WIDTH=169 HEIGHT=200 SRC="FIGDIR/small/727504v1_ufig1.gif" ALT="Figure 1"> View larger version (26K): org.highwire.dtl.DTLVardef@129eff8org.highwire.dtl.DTLVardef@49c401org.highwire.dtl.DTLVardef@e32b22org.highwire.dtl.DTLVardef@24f6df_HPS_FORMAT_FIGEXP M_FIG C_FIG

neuroscience↗

bioRxiv2026

SCALLOPS: a scalable, integrated computational framework for Optical Pooled Screens

Optical pooled screens (OPS) link pooled genetic perturbations to high-dimensional image-based phenotypes at scale, but their widespread adoption is hindered by computational bottlenecks in processing terabyte-scale, multimodal image data. We present SCALLOPS, a unified, modular, and cloud-native computational framework that overcomes these bottlenecks. SCALLOPS implements a "well-centric" processing strategy that integrates robust stitching with a non-linear two-stage registration strategy, enabling accurate alignment of multi-magnification images, reliable single-cell genotype-phenotype linkage, and efficient morphological feature extraction. Benchmarking with public and newly-generated datasets demonstrated SCALLOPS superior performance over existing solutions. Crucially, SCALLOPS uniquely enables robust processing of 4x magnification in situ sequencing data, accelerating image acquisition by around six-fold. We applied SCALLOPS to an optical pooled screen investigating the estrogen receptor (ER) degrader vepdegestrant in a breast cancer cell line, successfully recovering its known mechanism of action, highlighting the value of OPS in translational research. SCALLOPS provides a scalable end-to-end solution, making large-scale OPS routine.

bioinformatics↗

bioRxiv2026

A Geometric Model of Nucleus-Constrained Frustrated Phagocytosis

Frustrated phagocytosis occurs when phagocytes fail to fully engulf large targets, yet the geometric origins of this physical limit remain poorly defined. Here we present a geometric model that identifies the cell nucleus as an intracellular constraint on engulfment. Extending membrane-limited frameworks, we distinguish an intrinsic phagocytic capacity set by membrane availability from an apparent capacity reduced by nuclear exclusion. Using minimal geometric assumptions, we derive closed-form expressions linking experimentally measurable parameters, including target coverage, volume ratio, and size, to phagocytic capacity and a normalized axial separation that quantifies nuclear accommodation. The model predicts a size- and curvature-independent geometric criterion for nuclear involvement applicable to both spherical and planar targets. These results establish nuclear geometry as a fundamental physical bottleneck in phagocytosis and provide a quantitative framework for interpreting stalled engulfment and nuclear deformation-dependent responses.

cell biology↗

bioRxiv2026

Protenix-v2: Broadening the Reach of Structure Prediction and Biomolecular Design

Advances in biomolecular modeling have broadened the range of problems addressable by structure prediction and design models. Here, we present results from Protenix-v2, a system spanning high-accuracy structure prediction and biomolecular design. On the structure prediction side, Protenix-v2 achieves antibody-antigen success rates with up to 13-point gains over Protenix-v1, while 5-seed performance surpasses previous 1000-seed results. On the design side, Protenix-v2 demonstrates a 100% target-level success rate in novelty-controlled VHH-Fc campaigns, reaching hit rates up to 48%. Crucially, the model enables hit discovery on difficult GPCR targets with hit rates of 16%-88% (VHH-Fc) and up to 50% (mAb) under 16-30 testing budgets per target. Resulting hits show high developability and diversity. Beyond antibody tasks, we report improved ligand-related plausibility and successful cross-variant SARS-CoV-2 spike RBD mini-binder design. These results establish Protenix-v2 as a robust and powerful model for accelerated drug discovery.

molecular biology↗

bioRxiv2026

Geometry-enhanced protein language modeling enables discovery of novel antibiotic resistance genes

The global antibiotic resistome remains largely unexplored, not because antibiotic resistance genes (ARGs) are rare in the environment, but because many are evolutionarily distant from known ARGs. Current computational approaches primarily rely on sequence homology, and thus miss distant homologues. We develop GeoARG, a geometry-enhanced framework that integrates structural features with protein language models through knowledge distillation, enabling efficient large-scale screening using sequence input alone. Across multiple benchmarks, GeoARG substantially improves the detection of remotely homologous ARGs, particularly under low sequence identity and fragmented conditions. Large-scale metagenomic analysis uncovers 1,485 high-confidence ARG candidates that are highly divergent from known ARGs, expanding the phylogenetic and functional landscape of the resistome. Structural analyses further show that these candidates preserve active-site geometry and maintain stable ligand-binding configurations consistent with known resistance mechanisms. These results demonstrate that geometric constraints enable systematic expansion of the resistome and facilitate the discovery of evolutionarily distant yet functionally conserved genes. A public web server is available at https://ycclab.cuhk.edu.cn/GeoARG/.

bioinformatics↗

bioRxiv2026

Protenix-v1: Toward High-Accuracy Open-Source Biomolecular Structure Prediction

We introduce Protenix-v1 (PX-v1), the first fully open-source structure prediction model to attain superior performance to AlphaFold3 while strictly adhering to the same training data cutoff, model size, and inference budget. Beyond standard evaluations, we highlight the effectiveness of inference-time scaling behavior of Protenix-v1, demonstrating that increasing the sampling budget yields consistent improvements in prediction quality--a behavior previously observed in AlphaFold3 and largely absent from prior open-source models. In addition to improved accuracy, Protenix-v1 incorporates key capabilities including protein template integration and RNA MSA support. Furthermore, to better support real-world applications such as drug discovery, we additionally release Protenix-v1-20250630, a variant trained on a larger dataset (cutoff: June 30, 2025), delivering further improved prediction accuracy. Finally, we identify limitations in existing benchmarking practices and provide updated evaluation tools and year-stratified benchmarks to support more reliable and transparent assessment. Collectively, these contributions provide a robust foundation for the Protenix series and the broader field.

bioinformatics↗

bioRxiv2025

Multiple Molecular Pathways to Longevity: Opposing Gene Expression Programs Define Distinct Aging Strategies

While aging is the greatest risk factor for the development of neurodegenerative disease, the role of aging in these diseases is poorly understood. Our previous work has shown that targeting aging pathways can be neuroprotective in animal models of neurodegenerative disease. Based on these findings, we believe that by gaining insight into the aging process, that knowledge can be applied to identify novel therapeutic targets for neurodegenerative disease. To advance our understanding of aging, we used a genomics approach to identify genes regulated by multiple lifespan-extending pathways. We performed RNA sequencing on nine long-lived C. elegans mutants representing seven longevity pathways: insulin/IGF-1 signaling, dietary restriction, germline deficiency, impaired chemosensation, reduced translation, elevated mitochondrial ROS, and mild mitochondrial impairment. We found that most pairs of long-lived mutants exhibited a significant overlap in differentially expressed genes. Comparing gene expression across the entire panel of long-lived mutants revealed three distinct longevity groups that could be clearly distinguished by gene expression. Interestingly, two of these groups showed modulation of specific genetic pathways in opposite directions, suggesting that there are multiple alternative strategies to achieving long life. Filtering for genes similarly modulated in at least six mutants identified 196 upregulated and 62 downregulated aging genes. Upregulated genes were enriched in immunity, defense and metabolism, while many downregulated genes impacted translation and gene expression. To assess the ability of these genes to enhance longevity individually, we knocked down the commonly upregulated genes in long-lived mutants and evaluated the resulting effect on lifespan. Using this approach, we identified several genes that affect lifespan individually. Upregulation of at least some of these genes was sufficient to enhance stress resistance and extend lifespan in wild-type worms. Overall, the shared longevity genes identified in this work offer potential targets to promote healthy aging and decrease age-onset disease.

genetics↗

bioRxiv2025

Paralogs with distinct phase behaviors broaden the in vivo stress response range of condensates

Paralogs are widespread, but their physiological roles are often masked by redundancy. H-NS, a nucleoid-associated protein in Gram-negative bacteria, typically coexists with paralogs such as StpA, whose functions have remained obscured due to the lack of strong phenotypes. We demonstrate here that the interaction between H-NS and StpA fine-tunes the physico-chemical properties of heterochromatin-forming condensates. H-NS forms dynamic condensates, while StpA assembles into stable insoluble fibrils. Yet together, the two proteins form heterotypic condensates, whose fluidity and stability are tunable by their relative stoichiometry. By increasing the levels of StpA over H-NS, bacteria increase heterochromatin-associated condensate stability, thus preserving gene repression and optimizing bacterial growth under stress conditions. Structural differences at these proteins dimerization sites help explain their distinct phase behaviors. Our findings reveal a novel paradigm in which paralogs that are positioned at the opposite ends of the phase spectrum can fine-tune condensates to promote survival in fluctuating environments.

biochemistry↗

bioRxiv2025

Prion Seeding Activity in DNA Extractions: Implications for Laboratory Biosafety

Infectious prions (PrPSc) are largely resistant to proteolytic digestion, including proteinase K digestion. While nucleic acid extracts are generally considered non-infectious, we investigated whether standard DNA purification methods can co-purify PrPSc, posing an unrecognized biosafety risk. Two laboratories, the University of Minnesota Center for Prion Research and Outreach (MNPRO) and the Canadian Food Inspection Agency (CFIA), independently tested filter-based and magnetic bead-based DNA extraction kits using tissues from chronic wasting disease (CWD)-positive and -negative white-tailed deer (WTD; Odocoileus virginianus), as well as prion-infected and control Syrian hamster (Mesocricetus auratus) brains. CFIA used two filter-based kits (one automated and one manual), while MNPRO tested two manual kits (one filter-based and one magnetic bead-based). PrPSc seeding activity was measured in extracted DNA and source tissues using real-time quaking-induced conversion (RT-QuIC). MNPRO found substantial to almost perfect agreement (kappa ({kappa}) = 0.789 - 0.816) between RT-QuIC seeding activity of DNA eluates from both extraction methods and that of the source WTD tissue homogenate. CFIA optimized RT-QuIC to a 30-hour runtime, achieving 74% sensitivity and 94% specificity in 88 archived WTD DNA samples. Both laboratories concluded that commercial DNA extraction kits do not eliminate PrPSc, enabling its carry-over into DNA eluates. Until infectivity is resolved by animal bioassay, DNA from PrPSc-positive tissues should be managed under biosafety protocols appropriate for the originating prion disease, with appropriate decontamination and containment procedures.

molecular biology↗

bioRxiv2025

The neuropeptide neuromedin U receptor nmur-1 buffers insulin receptor signaling in bacteria-dependent C. elegans survival

Distinct microbial environments exert diverse effects on the physiology and survival of the nematode Caenorhabditis elegans. Here, we show that C. elegans grown on two Escherichia coli strains exhibit different survival dynamics. Wild-type C. elegans on the B type OP50 exhibit more early deaths compared to C. elegans on K-12 type CS180. These early deaths on OP50 are characterized by swollen pharynges (P-deaths) due to bacterial accumulation within the tissue. In contrast, animals on CS180 are more resistant to P-deaths. These bacteria-dependent differences in P-deaths depend on bacterial lipopolysaccharide structures and the activities of the C. elegans neuropeptide neuromedin U receptor NMUR-1, which reduces P-deaths on OP50, but not on CS180. Surprisingly, however, NMUR-1 promotes the opposite response when the insulin receptor DAF-2 has reduced function -- where NMUR-1 now stimulates P-deaths on OP50, but again with no effect on CS180. We also find that NMUR-1 acts in sensory neurons to promote its bi-directional effects on longevity, which depend on the FOXO transcription factor DAF-16. In addition, NMUR-1 downregulates the expression of the insulin-like peptide daf-28, but only when DAF-2 function is not reduced. This suggests a regulatory mechanism through which NMUR-1 maintains insulin receptor DAF-2 signaling at a suitable level. Thus, our studies reveal that NMUR-1 serves to buffer the dynamic range of DAF-2 receptor signaling, thereby optimizing pharyngeal health and survival in response to specific bacteria.

physiology↗

bioRxiv2025

PXDesign: Fast, Modular, and Accurate De Novo Design of Protein Binders

PXDesign achieves nanomolar binder hit rates of 17-82% across six of seven diverse protein targets, surpassing prior methods such as AlphaProteo. This experimental success rate is enabled by advances in both binder generation and filtering. We develop both a diffusion-based generative model (PXDesign-d) and a hallucination-based approach (PXDesign-h), each showing strong in silico performance that outperforms existing models. Beyond generation, we systematically analyze confidence-based filtering and ranking strategies from multiple structure predictors, comparing their accuracy, efficiency, and complementarity on datasets spanning de novo binders and mutagenesis. Finally, we validate the full design process experimentally, achieving high hit rates and multiple nanomolar binders. To support future research and broaden community adoption, we release the full PXDesign pipeline (https://github.com/bytedance/PXDesign), provide public access to PXDesign through a dedicated web server (https://protenix-server.com), and make all designed binder sequences available at the project page (https://protenix.github.io/pxdesign). O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=75 SRC="FIGDIR/small/670450v3_ufig1.gif" ALT="Figure 1"> View larger version (25K): org.highwire.dtl.DTLVardef@2a414aorg.highwire.dtl.DTLVardef@248f85org.highwire.dtl.DTLVardef@4a6afdorg.highwire.dtl.DTLVardef@1b622dd_HPS_FORMAT_FIGEXP M_FIG C_FIG

molecular biology↗

bioRxiv2025

ALK/ATR combination therapy is effective in neuroblastoma mouse tumors driven by MYCN

Summary paragraphOne of the key features of high-risk neuroblastoma (NB) is MYCN-amplification. While MYCN is still regarded as therapeutically challenging despite intensive efforts to find targeting compounds, inhibitors against Anaplastic Lymphoma Kinase (ALK) are now being evaluated among ALK-driven NB patients with promising results. There is a pressing need to find alternative treatment regimens for ALK naive MYCN-amplified patients, as current regimes are accompanied by significant mortality and morbidity. Here we show that genetically removing Alk in the Th-MYCN-driven engineered NB mouse model (GEMM) resulted in decreased tumor penetrance, and survival of Th-MYCN homozygote GEMMs, suggesting that Alk activity contributes to aggressiv e neuroblastoma development in this model. Given the high levels of replication stress induced in Th-MYCN tumors, we employed inhibitors of Ataxia Telangiectasia Rad3 related (ATR), combining ATR and ALK inhibition in a 14-day treatment regime, and comparing with ATR inhibitor monotreatment. Th-MYCN tumor bearing mice that received ATRi/ALKi combination treatment exhibited significantly increased survival compared to ATRi alone, that was sustained over time. Together, our data highlight a potentially effective treatment option for the currently challenging category of MYCN-amplified NB patients lacking ALK mutations.

cancer biology↗

bioRxiv2025

PlasRAG: comprehensive plasmid characterization and retrieval through sequence-text alignment

Plasmids play a pivotal role in the emergence of multidrug-resistant and pathogenic bacteria, posing significant clinical challenges. The integration of metagenomic sequencing with advanced bioinformatics tools surpasses traditional wet lab methods, leading to the discovery of millions of plasmids from diverse origins. However, the rapidly growing number of unannotated plasmids necessitates comprehensive characterization of their multi-faceted properties, such as risk indices and ecological contexts, to support various downstream applications. Achieving this goal is hindered by several challenges, including the limited availability of plasmid characterization tools, the inadequacies of alignment-based methods for novel plasmids, and inconsistencies in manual annotations across plasmid reference databases. To address these issues, we present PlasRAG, a novel tool that integrates two key modules: multi-faceted property characterization of query plasmids and plasmid DNA retrieval based on textual queries. At its core, PlasRAG employs a bidirectional multi-modal information retrieval model that aligns DNA sequences with textual data, effectively overcoming the limitations of traditional approaches. Specifically, within the characterization module, PlasRAG leverages the retrieval-augmented generation (RAG) framework and the Llama-3 large language model (LLM) to provide accurate and context-aware responses to user queries. Rigorous experiments demonstrate that PlasRAG delivers robust performance and enhanced analytical capabilities, underscoring the effectiveness of its architectural design. In particular, experiments on a real-world plasmid dataset curated from diverse human gut metagenomes suggest that plasmids with a broader host range and encoded ARGs tend to spread more extensively. The source code of PlasRAG is available via: https://github.com/Orin-beep/PlasRAG.

bioinformatics↗