Biology subjects

Zhang, G.

Publications and source records attributed to Zhang, G..

At least 127 records · Page 7Linked to original sources

bioRxiv2025

Revalidation of Manis aurita, based on integrative morphological and genomic evidence

Pangolins face severe conservation threats globally, necessitating accurate taxonomy for effective conservation. Previous genomic research on the Chinese pangolin (Manis pentadactyla) identified two deeply divergent lineages (MPA and MPB), suggesting underestimated species diversity. The recent description of M. indoburmanica (corresponding to MPB), however, did not assess its relationship with historical names, particularly the subspecies M. p. aurita, leaving the groups taxonomic status uncertain. To resolve this issue, we employed an integrative framework, analyzing genomic and morphological data from museum specimens including the lectotype of M. p. aurita to clarify phylogenetic relationships and taxonomy within M. pentadactyla sensu lato. Our results demonstrate that MPB, which includes both M. indoburmanica and M. p. aurita, is geographically restricted to the southern Himalayas and thus distinct from other M. pentadactyla populations. Genomic analyses indicate the two clades diverged [~]1.8 Ma and have remained largely isolated, with only minimal gene flow with congeners. Furthermore, morphometric analyses of both cranial and external features reveal consistent and significant differentiation between the Himalayan lineage (MPB) and M. pentadactyla sensu stricto. Collectively, these congruent findings provide unequivocal support for the revalidation of Manis aurita Hodgson, 1836, thereby establishing it as a distinct extant species of Asian pangolin.

evolutionary biology↗

bioRxiv2025

Integrative multi-omic analysis reveals oral microbiome-metabolome signatures of obesity

Obesity is a major global health challenge and a leading risk factor for cardiometabolic disorders. The global surge in obesity, driven by industrialization and the widespread consumption of low-fiber, ultra-processed food, highlights an urgent need for deeper biological insights. While the gut microbiome has been studied in the context of obesity, the contribution of the oral microbiome--the second largest microbial ecosystem in the human body-remains largely underexplored. Here, we report findings from a deeply-phenotyped prospective-cohort of 628 Emirati adults, leveraging amplicon sequencing of mouthwash samples and multi-omics profiling and functional and metabolic activity analysis of 97 obese individuals and 95 matched controls, making this the most comprehensive multi-omics analysis of the oral microbiome. We identified significant differences in oral microbial diversity, composition, functional pathways, and metabolic profiles between obese and non-obese groups. Integrated multi-omics analysis of the 192 matched metagenomes and metabolome samples uncovered significant metabolic reprogramming and altered energy regulation in obesity. Specifically, the oral microbiome of obese participants were enriched for the proinflammatory Streptococcus parasanguinis and Actinomyces oris, and the lactate-producing Oribacterium sinus. Many microbial pathways involved in dietary carbohydrate metabolism, histidine degradation, as well as the production of obesogenic biomolecules were also enriched in obese participants; however, B-vitamin and heme production pathways were depleted. Consequently, metabolites resulting from these pathways such as lactate, histidine derivatives, choline, uridine, and uracil were elevated in obesity. These consistent microbiome-metabolite shifts were strongly associated with prominent obesity-associated cardiometabolic markers, including serum triglycerides and alkaline phosphatases, establishing a robust link between oral microbiome and obesity. These findings provide the most comprehensive insights into how disrupted microbial-metabolic cross-talk in the oral cavity may contribute to obesity and related cardiometabolic disease risk, underscoring the potential of targeting of oral microbiome-host interactions as a novel avenue for obesity prevention and intervention.

microbiology↗

bioRxiv2025

Microglia surveillance is directed toward neuron activation during sustained intracortical microstimulation

Intracortical microstimulation (ICMS) is a widely used tool for neuroprostheses, but its long-term efficacy is often limited by biofouling and neuroinflammatory responses at the electrode-tissue interface. Microglia orchestrate neuroinflammation and regulate synaptic plasticity, and low-frequency stimulation has been shown to promote anti-inflammatory microglial phenotypes. We therefore examined, over the first three days post-implantation, how 10-Hz ICMS influences microglial-neuronal interactions in vivo using two-photon imaging in Cx3cr1-GFP/jRGECO1a mice. A one-hour session of 10-Hz ICMS did not induce overt morphological activation of microglia but increased their process motility, directing extensions toward both the electrode and neurons exhibiting elevated calcium activity. By post-implantation Day 2, microglial extensions were significantly biased toward neurons whose {Delta}F/F surpassed a 3 standard deviation threshold after stimulation onset (74.26{degrees} {+/-} 11.83{degrees}) but shifted away from those same neurons after 40 min of continuous stimulation (116.99{degrees} {+/-} 9.19{degrees}) (p = 0.001732), suggesting a dynamic, homeostatic response to sustained neuronal calcium elevations. Although multi-day electrode insertion accelerated microglial motility and polarization toward the device, 10-Hz ICMS alone did not alter microglial branching or soma shape. Microglial contact frequency scaled with neuronal adaptation profiles: depressed neurons received the most contacts immediately post-implant (1.15 {+/-} 0.3 contacts; p = .0460). These findings reveal stimulus-associated, neuron-dependent surveillance behaviors of microglia during early post-implantation ICMS and implicate them as active participants in short-term modulation of cortical circuits.

bioengineering↗

bioRxiv2025

Multimodal imaging reveals a lysosomal drug reservoir that drives heterogeneous distribution of PARP inhibitors

For all drugs, effective target engagement requires sufficient intracellular concentrations of drug to be reached, but whether tumour heterogeneity impacts drug distribution and efficacy is poorly studied. PARP inhibitors have transformed treatment of high-grade serous ovarian carcinoma (HGSOC), but resistance remains a clinical hurdle in this highly heterogeneous tumour type. We developed a patient-derived explant multi-modal imaging pipeline, which demonstrated that cell-intrinsic PARP inhibitor accumulation is highly variable, both between patients and within tumours. Spatial transcriptomics revealed enrichment of apoptotic and lysosomal signatures in high-drug regions. Rucaparib, an intrinsically fluorescent PARP inhibitor, accumulates heterogeneously at the single-cell level, with rucaparib-high cells demonstrating increased drug response relative to rucaparib low. Mechanistically, lysosomal sequestration creates a rucaparib reservoir that determines drug levels in the nucleus. Perturbation of lysosomal content altered intracellular levels of weak base PARP inhibitors rucaparib and niraparib, but not olaparib. Together these data suggest that lysosomes act as a reservoir for a subset of PARP inhibitor drugs to improve drug response.

cancer biology↗

bioRxiv2025

MultiSAAI: Sequence-informed Antibody-Antigen Interactions Prediction using Multi-scale Deep Learning

MotivationAntibodies possess high specificity toward target antigens, making them critical for therapeutic applications. However, experimental screening of antibody-antigen interactions is labor-intensive and costly. Existing structure-based computational methods often face significant challenges in modeling the dynamic conformational flexibility that governs antibody-antigen interactions, which limits their ability to accurately identify binding. Therefore, there is an urgent need for accurate and efficient sequence-based prediction method to accelerate antibody discovery and reduce experimental burden. ResultsWe propose MultiSAAI, a sequence-informed framework that models antibody-antigen interactions by explicitly accounting for the distinct roles of antibody heavy and light chains in antigen binding. MultiSAAI integrates language model embeddings, physicochemical properties, geometric constraints, and residue substitutability to characterize antibody-antigen interactions across multiple scales, employing a multi-scale network architecture that simultaneously evaluates global residue-pair compatibility and local amino acid fitness at the binding interface. Furthermore, the incorporation of site-specific information and biologically grounded binding principles allows the model to more closely reflect the actual mechanisms of interactions. Benchmark results demonstrate that MultiSAAI achieves AUROC scores of 0.757 on the generic antibody-antigen interactions dataset and 0.946 on the SARS-CoV-2 dataset, outperforming existing methods such as A2binder and AbAgIntPre. Finally, large-scale preliminary antibody screening further validates the potential of MultiSAAI for high-throughput therapeutic antibody discovery.

bioinformatics↗

bioRxiv2025

ZNF638 represses the transcription of HBV closed circular DNA involving HUSH complex-mediated histone modifications of epigenetic silencing

Hepatitis B virus (HBV) infection poses a critical threat to public health burden worldwide. As the major template of viral transcription, the covalently closed circular DNAs (cccDNAs) of HBV was known to form microchromosomes and interact with various host proteins. In the present study, ZNF638 was found to significantly reduce the transcription of HBV into pgRNAs, and subsequently reduce cccDNA formation and viral propagation. We discovered that ZNF638 was able to bind the preS and S gene regions of HBV, repressing the transcription of cccDNA upon increased modification of H3K9me3 mediated by SETDB1. This epigenetic silencing of cccDNA transcription was further demonstrated to involve the activity of HUSH, a multi-protein complex to mediate epigenetic modifications of histone into suppressed nucleosomes. In in vivo models, we evaluated the effects of ZNF638 deficiency on HBV-targeting siRNA therapy. A significant reduction of the efficacy was observed in animals with reduced ZNF638 expression. In summary, we report that ZNF638 is a potent host factor to repress the transcription of HBV DNAs and to attenuate HBV infection. The role of ZNF638 on HBV repression involves its binding to cccDNA at specific regions for recruiting HUSH complex to write H3K9me3 histone marks for epigenetic silencing. ZNF638 serves as a critical host factor to associate HBV infection and disease progression. Additionally, ZNF638 serves as a potential indicator to predict the outcome of HBV-targeting siRNA therapy.

biochemistry↗

bioRxiv2025

TransAgent: Dynamizing Transcriptional Regulation Analysis via Multi-omics-Aware AI Agent

Transcriptional regulation research, as a core area of life sciences, faces challenges such as scattered multi-omics data, complex joint analysis, and difficulties in integrating data processing tools. To address these issues, we propose TransAgent, an agent software specifically designed for transcriptional regulation analysis. Through innovative designs such as multi-mode operation (planning/execution/automatic), dynamic memory management, rapid MCP tool expansion (integrating over 30 tools), integration of transcriptional regulation annotation data (over 20 data sources including epigenomics and gene expression profiles), and cloud Docker computing, TransAgent significantly improves analysis efficiency. We have successfully applied TransAgent to various transcriptional regulation analysis scenarios such as re-construction of super-enhancer regulatory circuit in esophageal squamous cell carcinoma and identification of key regulators in cardiomyocyte differentiation, demonstrating analytical robustness and uncovering biological insights. TransAgent automates the entire process from raw data processing to advanced analysis, such as joint prediction of multi-omics data, transforming traditionally time-consuming and labor-intensive tasks into a conversation-driven approach. This provides a new paradigm for transcriptional regulation research, centered around large models as the core driver of scalable agent application analysis.

bioinformatics↗

bioRxiv2025

Evolution of a fuzzy ribonucleoprotein complex in viral assembly

Previously we showed that the genetic diversity of SARS-CoV-2 nucleocapsid (N) protein explores a wide range of biophysical properties facilitated by non-local impact of point mutations to its intrinsically disordered regions (Nguyen et al., 2024). This includes modulation of self-association, such as the creation of a de novo binding interface through the P13L mutation characteristic of Omicron variants. In the present work we focus on the key function of N condensing viral RNA into ribonucleoprotein particles (RNPs) for viral assembly. Lacking high-resolution structural information, biochemical and biophysical approaches have revealed architectural principles of RNPs, which involve cooperative interactions of several protein-protein and protein-RNA interfaces, initiated through oligomerization of conserved transient helices in the central disordered linker of N. Here we study the impact of defining N-protein mutations in variants of concern on RNP formation, using biophysical tools, a virus-like particle assay, and reverse genetics experiments. We find convergent evolution in repeated, independent introduction of amino acid substitutions strengthening existing binding interfaces, compensating for other substitutions that promote viral replication but decrease RNP stability. Furthermore, we show that the P13L mutation of Omicron variants enhances RNP assembly and increases viral fitness. Overall, our data reveal RNP complexes to be highly variable not only in sequence and conformations, but also in thermodynamic and kinetic stability, with its pleomorphism affecting basic architectural principles. We hypothesize that the formation of polydisperse, fuzzy N-RNA clusters with multiple distributed weak binding interfaces optimizes reversible RNA condensation, while supporting host adaptation and allowing for a large sequence space to be explored.

biophysics↗

bioRxiv2025

Enhanced protein precipitation with ammonia enables rapid, universal extraction of oligonucleotides for bioanalysis

Extracting oligonucleotides from biological matrices for mass spectrometry (MS) using current methodologies is time-consuming and costly. Accelerating the discovery, characterization, and development of oligonucleotides (ONTs) requires a universal method that is low-cost, efficient, without requiring extensive method development for each new test article. Protein precipitation using organic solvents meets these criteria for small molecules but has historically failed for ONTs due to poor recovery arising from their tendency to co-precipitate with proteins. This study introduces an effective approach utilizing small amines dissolved in organic solvents to significantly boost extraction recovery of ONTs from biological matrices .We first present the methods development and then analytically qualify it for the bioanalysis of antisense oligonucleotides (ASOs) and small interfering RNAs (siRNAs) extracted from various tissues, using ion-pairing reverse phase (IPRP) liquid chromatography coupled to tandem MS and high resolution MS (HRMS) detection. This method, Enhanced Protein Precipitation (EPP), achieves near-complete recovery across multiple ONT classes from biological matrices without need of sample digestion, costly solid phase extraction plates, or custom-designed hybridization probes. The method has proven to be more versatile and sustainable than conventional approaches. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=118 SRC="FIGDIR/small/650521v2_ufig1.gif" ALT="Figure 1"> View larger version (15K): org.highwire.dtl.DTLVardef@a623d2org.highwire.dtl.DTLVardef@179b053org.highwire.dtl.DTLVardef@27273corg.highwire.dtl.DTLVardef@80fe47_HPS_FORMAT_FIGEXP M_FIG C_FIG

biochemistry↗

bioRxiv2025

When discrete characters are wanting: Continuous character integration under the phylospecies concept informs the revision of the Australian land snail Thersites (Eupulmonata, Camaenidae)

Species that are predominantly characterized by continuous instead of discrete morphological characters pose a challenge to species delimitation. Under the phylospecies concept, species are delimited by apomorphies, which are difficult to establish when characters are not discrete. In the present study, we address these challenges in the Australian land snail genus Thersites (family Camaenidae), where accepted species exhibit a scarcity of discrete distinguishing characters and differ in a limited range of continuous characters according to prior taxonomic studies. We integrate analyses of genome-scale molecular data with evaluations of several continuous, qualitative, and discrete morphological characters derived from landmarks to delineate species by identifying apomorphies. We found that dimensions of shell and genitals overlapped considerably among the currently accepted species. These overlaps may indicate morphospace saturation, which may affect species delineation through the gen-morph species concept. Statistical methods, such as Dunns test, failed to consistently delineate monophyletic taxa. Additionally, we could not derive apomorphies from discretized landmarks for the prevalence of outlier specimens. However, applying the Kruskal-Wallis test at certain nodes of the tree revealed significant differences in some continuous characters. We propose that these inferred differences represent apomorphies of taxonomic lineages. Ultimately, we suggest based on our findings that of the four currently accepted species, two should be synonymized. To maintain monophyly of taxa, we synonymize Thersites mitchellae with T. novaehollandiae and T. darlingtoni with T. richmondiana. Distinct character states of the umbilicus support the existence of two independent lineages: One with an open (T. sp1 + T. sp2) and one with a closed umbilicus (T. novaehollandiae + T. richmondiana). Overall, the Kruskal-Wallis tests support four distinct species, T. novaehollandiae and T. richmondiana plus two newly described species. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=122 SRC="FIGDIR/small/649343v2_ufig1.gif" ALT="Figure 1"> View larger version (58K): org.highwire.dtl.DTLVardef@1d9dc53org.highwire.dtl.DTLVardef@ff619org.highwire.dtl.DTLVardef@1e618aeorg.highwire.dtl.DTLVardef@15e80e6_HPS_FORMAT_FIGEXP M_FIG C_FIG

evolutionary biology↗

bioRxiv2025

High-accuracy protein complex structure modeling based on sequence-derived structure complementarity

In living organisms, proteins perform key functions required for life activities by interacting to form complexes. Determining the protein complex structure is crucial for understanding and mastering biological functions. Although AlphaFold2 has made a revolutionary breakthrough in predicting protein monomeric structures, accurately capturing inter-chain interaction signals and modeling the structures of protein complexes remain a formidable challenge. In this work, we report DeepSCFold, a pipeline for improving protein complex structure modeling. DeepSCFold uses sequence-based deep learning models to predict protein-protein structural similarity and interaction probability, providing a foundation for identifying interaction partners and constructing deep paired multiple-sequence alignments (MSAs) for protein complex structure prediction. Benchmark results show that DeepSCFold significantly increases the accuracy of protein complex structure prediction compared with state-of-the-art methods. For multimer targets from CASP15, DeepSCFold achieves an improvement of 11.6% and 10.3% in TM-score compared to AlphaFold-Multimer and AlphaFold3, respectively. Furthermore, when applied to antibody-antigen complexes from the SAbDab database, DeepSCFold enhances the prediction success rate for antibody-antigen binding interfaces by 24.7% and 12.4% over AlphaFold-Multimer and AlphaFold3, respectively. These results demonstrate that DeepSCFold effectively captures intrinsic and conserved protein-protein interaction patterns through sequence-derived structure-aware information, rather than relying solely on sequence-level co-evolutionary signals.

bioinformatics↗

bioRxiv2025

On the impact of reference selection on variant calling in phylogenomics: How to avoid systematic error in target-enrichment studies of non-model organisms

Target-enrichment methods are widely used in phylogenomic studies aiming to reconstruct the relationships among diverse groups of organisms. When building phylogenomic datasets from short sequencer reads, software pipelines need to accurately align short sequencer reads to the target and identify the genetic variation in homologous DNA sequences. Variant calling, a crucial step in this process, is known to be error-prone in a phylogenetic analysis framework as it requires a conspecific reference as a template. To avoid problems relating to reference choice, existing target-enrichment pipelines often skip variant calling entirely using work-arounds. This study investigates the effect of reference genome selection on the reliability of variant calling and assesses the consequences of reference genome choice on phylogenomic analyses downstream. Employing an empirical exon capture sequence dataset, we examine how variant detection is influenced by reference genome choice. We reconstructed and analysed phylogenomic sequence datasets from short sequencer reads, which includes steps of quality control, read mapping, variant calling, multiple sequence alignment and phylogenetic analysis. We assembled the same exon capture dataset by using four different references for variant calling. These references were (1) a distantly related species, (2) a single ingroup sample, (3) a composite of all ingroup samples, and (4) a self-derived reference for each sample. We found that reference choice significantly impacted the variant detection and that these differences in variant detection influenced the phylogenetic reconstructions downstream. Comparing multiple sequence alignments and phylogenetic trees produced using different references revealed that employing a sample-specific self-reference for variant calling maximizes the accuracy of the variant detection process. Based on this finding, we recommend incorporating self-referenced variant calling in the assembly of phylogenomic datasets to ensure robustness and reproducibility, especially for datasets characterized by low read coverage.

evolutionary biology↗

bioRxiv2025

High-efficiency likelihood inference of shared proteomic architectures across 50 complex human traits

Advancements in genetic correlation estimation have elucidated genome-wide pleiotropys influence on phenotypic correlations among human complex traits and diseases. However, the role of proteomic domains in these correlations remains underexplored. Traditional genetic correlation analysis assumptions, including the minute effects of SNPs and their linkage disequilibrium, do not suit proteomic data. We present a novel method, Likelihood-based Estimation for Proteomic Correlation (LEAP), tailored to provide unbiased estimation of shared proteomic architectures between trait pairs. LEAP notably decreases computational demands by approximately 1000-fold compared to conventional bivariate linear mixed models. We applied LEAP to data from the UK Biobank Pharma Proteomics Project, identifying 585 significant proteomic correlations among 1,225 pairs of 50 biochemical, anthropometric, and behavioral traits. Furthermore, we quantified the distinct proteomic and genetic contributions to phenotypic correlations, highlighting significant gender differences. This study provides a comprehensive computational approach for proteomic correlation estimation, clarifying the specific roles of genomics and proteomics in complex trait correlations. Our findings not only advance the understanding of proteomic contributions to phenotypic traits but also suggest potential applications for evaluating shared omics architectures in other domains such as transcriptomics and metabolomics.

bioinformatics↗

bioRxiv2025

AGP-Net: A Universal Network for Gene Expression Prediction of Spatial Transcriptomics

In the era of high-throughput biology, molecular phenotypes have proven effective in predicting disease states and future trajectories. Transcriptomics, in particular, has enabled the dissection of complex diseases with heterogeneous genetic and environmental aetiology, both aiding diagnosis and augmenting treatment. As improving technology has led to measurements of gene expression at increasing granularity, it has become progressively feasible to resolve disease traits that present locally or with spatial heterogeneity. Principal among these are cancers in which tumour gene expression, while itself heterogenous, exhibits a distinct signature from that of the surrounding tissue. Identifying this signature through molecular phenotyping facilitates specific cancer diagnosis and treatment. Here, we introduce AGP-Net, a multi-modal foundation framework capable of predicting gene expression from histopathology images. Rather than produce an aggregated estimate of expression for each gene, AGP-Net disaggregates images into spots and attempts to resolve the variation in gene expression across them, thereby providing coarse spatial transcriptomic predictions across the tumour slice and surrounding region. The challenge in doing so is due to data sparsity relative to the dimensionality of the problem: the number of genes and their contextual heterogeneity within and between tissue and cancer types makes it difficult to train a model on the limited data available. The innovation of AGP-Net lies in borrowing strength across similar genes as defined by their textual language descriptions. Our AGP-Net supports datasets with varying gene coverage and facilitates the prediction of gene expression for previously unseen genes based on their textual descriptions. Trained on millions of spots from diverse dataset sources, AGP-Net establishes state-of-the-art performance in zero-shot spatial gene expression prediction, demonstrating its adaptability to generalize across novel scenarios.

pathology↗

bioRxiv2025

KcatNet: advancing genome-wide enzyme turnover number prediction through structural enzymatic characterization

Enzyme turnover numbers (Kcat) are fundamental kinetic constants that quantify enzymatic efficiency. Systematic studies of Kcat are essential for characterizing the mechanisms underlying proteomic composition and cellular metabolism. However, experimental measurements of Kcat remain limited and prone to noise. To address this, we present KcatNet, a geometric deep learning model designed for high-throughput prediction of Kcat in metabolic enzymes across all organisms, leveraging paired enzyme sequence and substrate representations. KcatNet consistently outperforms existing predictors, particularly for enzymes with high catalytic efficiency, and demonstrates strong generalization to enzymes that are dissimilar to those in the training set. Furthermore, KcatNet uncovers structural mechanisms and interaction patterns within enzyme-substrate complexes, providing insights into architectural principles that are inaccessible with existing methods by harnessing the representational power of large-scale protein language models. We applied KcatNet to genome-scale Kcat prediction across diverse yeast species, improving proteome allocation predictions by integrating its outputs into metabolic models. Experimental validation further confirmed the models ability to identify enzyme mutants with enhanced activity. By bridging the gap between sequence, structure, and function, KcatNet establishes a robust foundation for advancing understanding of molecular-level mechanisms and accelerating enzyme engineering efforts.

bioinformatics↗

bioRxiv2025

Inwardly rectifying potassium channels regulate membrane potential polarization and direction sensing during neutrophil chemotaxis

Potassium channels regulate membrane potential and diverse physiological processes, including cell migration. However, the specific function of the inwardly rectifying potassium channels in immune cell chemotaxis is unknown. Here, we identified that the inwardly rectifying potassium channel Kir7.1 (KCNJ13) maintains the resting membrane potential and is required for directional sensing during neutrophil chemotaxis. Pharmacological or genetic inhibition of Kir7.1 in neutrophils impaired direction sensing toward various chemoattractants without affecting cell polarization in multiple neutrophil models. Using genetically encoded voltage indicators, we observed oscillating depolarization of the membrane potential in protrusions in zebrafish neutrophils, and Kir7.1 is required for polarized depolarization towards the chemokine source. Focal depolarization with optogenetic tools biases pseudopod selection and induces de novo protrusions. Global hyperpolarizing neutrophils stalled cell migration. Furthermore, Kir7.1 regulates GPCR signaling activation. This work adds membrane potential to the intricate feedforward mechanism, coupling the adaptive and excitable network required to steer immune cells in complex tissue environments.

cell biology↗

bioRxiv2025

DeepSeMS: a large language model reveals hidden biosynthetic potential of the global ocean microbiome

Microbial biosynthetic diversity holds immense potential for discovering natural products with therapeutic applications, yet a substantial quantity of natural products derived from uncultivated microorganisms remains uncharacterized. The intricate nature of biosynthetic enzymes poses a major challenge in accurately predicting the chemical structures of secondary metabolites solely based on genome sequences using current rule-based methods. Here, we present DeepSeMS, a large language model designed to predict the chemical structures of secondary metabolites from various microbial biosynthetic gene clusters. Built on the Transformer architecture, DeepSeMS innovatively identifies sequence features using functional domains of biosynthetic enzymes, and incorporates feature-aligned chemical structure enumeration for training data augmentation. External evaluation results show that DeepSeMS predicts more accurate chemical structures of secondary metabolites with a Tanimoto coefficient up to 0.6 compared with the ground truth, significantly outperforming antiSMASH and PRISM with coefficients of only 0.14 and 0.45 respectively. Moreover, DeepSeMS successfully predicted secondary metabolites for 96.60% of cryptic biosynthetic gene clusters, surpassing existing methods with success rates less than 50%. Leveraging DeepSeMS, we characterized over 65,000 novel secondary metabolites from the global ocean microbiome with previously undocumented structural types, ecological distribution, and biomedical applications especially antibiotics. A login-free and user-friendly web server for DeepSeMS (https://biochemai.cstspace.cn/deepsems/) has been launched, featuring an integrated global ocean microbial secondary metabolites repository to expedite the discovery of novel natural products. Collectively, this study underscores the great capacity of a large language model-driven method in revealing hidden biosynthetic potential of the global ocean microbiome.

synthetic biology↗

bioRxiv2025

Trimethylamine-N-oxide affects cell type-specific pathways and networks in mouse aorta to promote atherosclerotic plaque vulnerability

BackgroundTrimethylamine-N-oxide (TMAO) has been significantly linked to atherosclerosis via several mechanisms, but its direct effect on the atherosclerosis-prone vasculature remains unclear. The objective of this study was to characterize the cell type-dependent and independent effects of TMAO on key vascular cell types involved in atherosclerosis progression in vivo. MethodsWe performed single cell RNA-sequencing (scRNAseq) on aortic athero-prone regions of female Ldlr-/- mice fed control Chow, high-cholesterol (HC), or HC+TMAO diets for three months to identify which aortic cell types, differentially expressed genes, and biological pathways are affected by TMAO. We also modeled cell-cell communications and intracellular gene regulatory networks to identify gene networks perturbed by TMAO feeding. Key genes and pathways were validated using primary human smooth muscle cells exposed to TMAO. Changes in the thickness of lesional fibrous caps in response to TMAO in female Ldlr-/- mice fed HC+TMAO versus HC diets were measured using transgelin immunostaining. ResultsOur scRNAseq analysis revealed that TMAO supplementation upregulated apoptotic gene signatures and downregulated extracellular matrix (ECM) organization and collagen formation genes in a subset of atherosclerosis-specific modulated vascular smooth muscle cells (vSMCs). We also identified "degradation of the ECM" as a top pathway for SMC-derived macrophage DEGs in response to TMAO. Network analyses support that macrophage-vSMC communication mediates ECM remodeling. Using human smooth muscle cells exposed to TMAO in vitro, we confirmed the direct effect of TMAO on regulating collagen and apoptotic genes. In agreement with the changes in these pathways that affect plaque stability, we observed a significant decrease in fibrous cap thickness in mice supplemented with TMAO. ConclusionsOur results reveal the effects of TMAO on vSMCs to promote apoptosis and decrease ECM formation, and on macrophage-mediated ECM degradation in atherosclerotic lesions to in concert enhance plaque instability. Graphic Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=84 SRC="FIGDIR/small/640205v1_ufig1.gif" ALT="Figure 1"> View larger version (33K): org.highwire.dtl.DTLVardef@100254org.highwire.dtl.DTLVardef@109155aorg.highwire.dtl.DTLVardef@158d280org.highwire.dtl.DTLVardef@63d544_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsO_LIscRNAseq of the aortic athero-prone regions in female Ldlr-/- mice supplemented with TMAO in the diet revealed the effect of TMAO across cell types, particularly in SMC-derived macrophages and atheroprotective modulated vSMCs. C_LIO_LITMAO increases apoptotic gene signatures and reduces ECM organization and collagen formation gene signatures in modulated vSMCs in vivo, and in vitro exposure studies support a direct effect of TMAO on these genes. C_LIO_LIModulated vSMC-specific gene regulatory networks enriched for apoptotic genes and ECM organization genes were organized by intracellular regulators such as Ccl19 and Tnn and extracellular regulators such as Mmp9 and Spp1 from macrophages. C_LIO_LIFibrous cap thickness, a marker of atherosclerotic plaque stability, was significantly reduced in female Ldlr-/- mice fed HC+TMAO versus HC diets for five months. C_LI

physiology↗