Biology subjects

Xie, P.

Publications and source records attributed to Xie, P..

At least 19 recordsLinked to original sources

bioRxiv2026

Charting sex divergence in the human brain across the lifespan

Biological sex fundamentally shapes human brain organization, but sex-specific normative neuroanatomical trajectories across the lifespan remain largely uncharted. Here, we constructed independent, sex-specific lifespan brain charts using structural neuroimaging data from 59,915 healthy individuals (29,760 males and 30,155 females) ranging in age from 266 postconception days to 100 years. By examining 296 structural phenotypes across global, cortical, and subcortical measures, these models revealed widespread sex differences in maturational timing, with males reaching peak milestones later than females. These trajectories demonstrate that sex differences evolve dynamically, with phenotype-specific windows of emergence and maximal separation. Compared with conventional sex-pooled references, sex-specific models achieved superior predictive accuracy and reduced misestimation of individual deviations in healthy populations. Across five neuropsychiatric conditions, sex-specific models improved the detection of extreme deviations and revealed both shared and sex-dependent patterns of disorder-related neuroanatomical abnormalities. These sex-specific charts establish tailored normative references for assessing brain development, ageing, and disease.

neuroscience↗

bioRxiv2026

The cellular and genetic basis of inflorescence divergence between maize and teosinte

The domestication of maize from teosinte involved dramatic remodeling of the ear, yet the cellular and genetic bases of this transformation remain unclear. Here, we generate a single-nucleus and spatial transcriptome atlas of developing maize and teosinte ears. Comparative analysis reveals divergence in cob-associated cell types, with enhanced cytokinin signaling and reduced growth-inhibitory signals collectively driving cob thickening and enlargement in maize. We further demonstrate that domestication expanded the spatial expression domain of key transcription factors in maize meristem cells, enhancing the potential for increasing kernel number. Additionally, we verified a major domestication gene, ZmSPD1, in which two nonsynonymous SNPs differentiate maize from teosinte and alter jasmonic acid (JA) levels in the ear, thereby suppressing spikelet abortion to effectively double kernel production. These findings provide a cell-resolved mechanistic framework for how cob architecture and kernel number were shaped during maize domestication, offering new insights into the formation of key agronomic traits.

plant biology↗

bioRxiv2026

Non-Invasive Embryo Quality Assessment via Matrix-Optimized Untargeted LC-MS Metabolomics of Spent Embryo Culture Media and Weighted Ensemble Machine Learning

BackgroundNon-invasive embryo quality assessment is a critical unmet need in assisted reproductive technology (ART). Preimplantation genetic testing for aneuploidy (PGT-A) is effective but requires invasive biopsy that may compromise embryo viability. Metabolomics of spent embryo culture media (SECM) offers a non-invasive alternative, yet analytical challenges--limited sample volume, high salt content, and abundant proteins--have hindered standardization and clinical translation. ResultsWe systematically optimized sample preparation for untargeted LC-MS metabolomics of SECM using human serum as a reference. Optimal conditions were highly matrix-dependent: SECM required 7x volume of 50% acetonitrile for extraction and 40% acetonitrile for reconstitution, whereas serum required 10x volume of 100% methanol and 100% water--reflecting that SECM contains more non-polar species than serum. Applying the optimized workflow to 120 clinical SECM samples (72 euploid, 48 aneuploid), we identified 102 differential metabolites between euploid and aneuploid embryos, with prominent enrichment of lipid pathways (fatty acid metabolism, {beta}-oxidation, sphingolipid metabolism) and involvement of amino acid (methionine, tryptophan) and TCA cycle metabolism. A weighted ensemble machine learning model discriminated aneuploid from euploid embryos with an AUC of 0.977, 100.0% specificity, and 89.6% sensitivity. Among 72 euploid embryos stratified by morphological grading (good, fair, poor), metabolic alterations progressed from mitochondrial energy deficiency (good vs. fair) to broader lipid dysregulation (fair vs. poor), with the ensemble model achieving AUCs of 0.944, 0.889, and 0.943, respectively. ConclusionsThis study establishes a rigorously optimized and validated SECM metabolomics workflow that overcomes key analytical barriers in this challenging matrix. Our findings demonstrate that metabolic signatures--particularly in lipid and energy metabolism--are strongly associated with both embryo ploidy and morphological quality, providing biological insights into the metabolic underpinnings of embryo developmental competence. The high predictive performance of the ensemble model supports the feasibility of non-invasive embryo assessment as a complementary tool to existing methods, with potential to reduce reliance on invasive biopsy in ART. External validation in prospective multi-center cohorts is warranted to further assess clinical utility and generalizability. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=122 SRC="FIGDIR/small/741666v1_ufig1.gif" ALT="Figure 1"> View larger version (59K): org.highwire.dtl.DTLVardef@10fbe40org.highwire.dtl.DTLVardef@936273org.highwire.dtl.DTLVardef@246e5dorg.highwire.dtl.DTLVardef@1c9d115_HPS_FORMAT_FIGEXP M_FIG C_FIG

genetics↗

bioRxiv2026

Apollo 3: Multi-Species Genome Curation

We present Apollo 3, a new manual genome annotation tool that integrates with the JBrowse 2 genome browser. Its functionality is inspired by existing manual genome annotation tools such as Apollo, Artemis, and Otter, but uses an updated and more scalable architecture and technology stack. It allows the simultaneous editing of multiple genomes, including the visualization of synteny to inform those annotations. Apollo 3 can be used as a standalone annotation editor, or it can be installed on a server and used collaboratively. We describe the application's design, features, and use cases.

genomics↗

bioRxiv2026

Large Language Model Agent Enables Autonomous Machine Learning Model Building for Biomedicine

Machine learning accelerates biomedical discovery, but creating effective predictive models requires specialized human expertise and demanding manual effort. Researchers must iteratively design pipelines, select architectures, and debug code. This challenge is particularly severe in biomedicine because of the heterogeneous datasets, sparse annotations, and complex evaluation protocols that are common in the domain. We present AIDO.Builder, an agentic artificial intelligence system that fully automates the entire life-cycle of biomedical model development. Provided only with a natural language task description and a target metric, AIDO.Builder autonomously constructs executable training and evaluation pipelines. The system selects suitable modeling strategies, executes experiments, and uses automated feedback-loop to iteratively revise its own code, configurations, and training procedures. It flexibly adapts to new tasks by training specialized models de novo or by using pretrained foundation models to build predictive models through task-appropriate adaptation. We show that across diverse biomedical benchmarks, AIDO. Builder produces highly competitive solutions against human alternatives, while eliminating the manual iteration previously required for robust model development. By automating the translation of raw data into reliable AI models, AIDO.Builder demonstrates how AI itself can be used to accelerate AI for biomedical research.

bioinformatics↗

bioRxiv2026

Harnessing AI to Build Virtual Cells

A virtual cell is a world model of a cell: a computational system that predicts, simulates and programs cellular processes across modalities and scales. An important path toward this goal is to model how genetic and chemical perturbations give rise to transcriptional responses, a core capability for disease understanding and drug discovery. However, current approaches remain expert-intensive, relying on iterative manual model design, training and debugging over months. Here we present VCHarness, an autonomous AI system that constructs perturbation-response models by combining an AI coding agent with multimodal biological foundation models. The system explores large spaces of architectures and training pipelines with minimal human intervention, iteratively generating, evaluating and refining candidate models. Across multiple perturbation-response benchmarks, VCHarness identifies architectures that outperform expert-designed approaches while reducing development time from months to days. It further uncovers non-obvious architectural patterns associated with improved performance, indicating that automated search can extend beyond conventional design strategies. These results suggest a shift from manually engineered models toward autonomous systems for constructing components of virtual cell world models, enabling scalable and data-driven exploration of cellular systems.

bioinformatics↗

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

OSTM1 is a ubiquitin E3 ligase that suppresses B-cell malignancy by activating the cAMP/PKA/CREB pathway

Osteoclastogenesis-associated transmembrane protein 1 (OSTM1) is a glycosylated, membrane-integral protein that regulates lysosomal homeostasis, with loss-of-function mutations causing autosomal recessive osteopetrosis. Through a whole-genome CRISPR/Cas9 screen, we identified OSTM1 as a previously unrecognized tumor suppressor in B-cell malignancies. Consistent with this role, OSTM1 is frequently deleted or downregulated across a broad spectrum of human B-cell cancers. In mice, B-cell-specific monoallelic or biallelic ablation of Ostm1 cooperates with Cdkn2a loss to drive lymphomagenesis with near-complete penetrance. Mechanistically, we uncover a cytosolic, non-glycosylated fraction of OSTM1 that functions as a ubiquitin E3 ligase to promote proteasomal degradation of phosphodiesterase 3B (PDE3B). As PDE3B hydrolyzes cAMP and suppresses the tumor-protective PKA/CREB/CREBBP signaling axis, loss of OSTM1 leads to PDE3B stabilization, attenuation of cAMP signaling, and enhanced oncogenic transformation. Together, our findings establish OSTM1 as a critical suppressor of B-cell lymphomagenesis through regulation of the cAMP/PKA/CREB pathway.

cancer biology↗

bioRxiv2025

MetaboliteChat: A Unified Multimodal Large Language Model for Interactive Metabolite Analysis and Functional Insights

Accurately characterizing the mechanisms and properties of metabolite molecules is essential for advancing metabolomics and systems biology. Yet most existing approaches are narrow, task-specific models that do not transfer across tasks and cannot express the richer, multimodal biological context of a metabolite in natural language. To address these challenges, we introduce MetaboliteChat, a multimodal ChatGPT-like large language model (LLM) that provides a unified and interactive framework for metabolite analysis. MetaboliteChat integrates molecular-graph and image reasoning with natural language understanding to generate comprehensive, free-form predictions about metabolite mechanisms and properties. The MetaboliteChat architecture consists of a graph neural network (GNN), a convolutional neural network (CNN), a large language model (LLM), and adapters, all trained end to end. This unified, multimodal design enables interactive reasoning over unseen metabolites, allowing the model to integrate structural and contextual cues and support discovery and translational insights across diverse biological systems.

bioinformatics↗

bioRxiv2025

T Cells Tear Apart Confining Extracellular Matrix Via a Breaststroke-like Motion to Generate Migration Paths

T cells adeptly migrate through soft tissues to target aberrant cells and regulate immunity. However, how they establish migration paths in confining nanoporous extracellular matrices (ECMs), and why they often fail to do so in dense ECMs that occur during fibrosis and around tumors, remain unclear. Here, we studied T cell migration in confining collagen-rich hydrogels spanning a range of stiffness, viscoelasticity, mechanical plasticity, and shear strength. Strikingly, only shear strength--the stress required for material failure--correlated strongly with migration, challenging the long-held focus on stiffness and pore size in cell motility. During migration, T cells extend actin-rich, finger-like protrusions into the ECM, which then undergo divergent breaststroke-like motion. Thus, T cells tear apart confining matrices using breaststroke-like motion to generate migration paths.

biophysics↗

bioRxiv2025

RobPicker: A Meta Learning Framework for Robust Identification of Macromolecules in Cryo-Electron Tomograms

Accurate particle picking of macromolecules of all kinds, shapes, and sizes in cryogenic electron tomograms (cryo-ET) is critical for understanding the molecular architecture of biological systems in their native state. Current deep learning methods have shown potential in identifying macromolecules from tomograms, but they are vulnerable to issues of noise in the training dataset obtained through human or automatic labeling and imbalanced distribution of macromolecule species. To address these limitations, we developed RobPicker, a meta-learning framework that effectively mitigates these issues by automatically learning deep neural networks to correct label errors and give greater emphasis to underrepresented macromolecule species. In evaluations across diverse cryo-ET datasets with noisy labels and imbalanced species distributions, RobPicker substantially outperforms state-of-the-art methods, particularly in identifying small and rare macromolecules. The efficiency and robustness of RobPicker can also be used for rapid fine-tuning of the tilt-series alignment, leading to improved tomogram reconstruction and enabling high-resolution cellular structural biology analysis.

bioinformatics↗

bioRxiv2025

Rapid, Efficient Candida albicans Lysis Comparison Using the Cellometer X2 Fluorescent Cell Viability Counter

Cell lysis is essential for extracting intake genetic material, forming the basis for diagnostic tests, and genetic studies. Commonly used lysis methods include thermal lysis, mechanical force, chemicals, biologicals, and sonication. Determining effective lysis methods for specific cell types is crucial for conducting further research. This study evaluates the lysis efficiency of Candida albicans using the Cellometer X2 fluorescent cell viability counter, employing various lysis methods: thermal, chemical, and enzymatic. Our results indicate that high-pH buffers combined with heat treatment enhance lysis efficiency, while SDS alone or with Proteinase K is insufficient for lysis at room temperature. In contrast, Zymolyase effectively lyses C. albicans at room temperature in 150 minutes, making the automation of room temperature lysis and nucleic acid purification feasible for C. albicans. Overall, this study highlights the Cellometer X2s capability for rapid and direct evaluation of cell lysis efficiency.

microbiology↗

bioRxiv2025

Sephin1 alleviates white matter injury by protecting oligodendrocyte after intracerebral hemorrhage

BackgroundWhite matter injury (WMI) caused by intracerebral hemorrhage (ICH) is a major neuropathological feature closely associated with neurological impairments such as motor and sensory dysfunction. Oligodendrocytes (OLs), which are responsible for repairing WMI, also suffer severe death resulting from the compression of hematoma and secondary neuroinflammation after ICH. Sephin1, a selective inhibitor of PPP1R15A, has been shown to reduce general protein synthesis and protect OLs by prolonging the integrated stress response (ISR). We aimed to evaluate the effectiveness of Sephin1 in protecting OLs in experimental ICH mice and primary OLs and microglia co-cultures. MethodsWe first determined the performance of ICH mice treated with Sephin1 or vehicle in multiple behavioral tests. To investigate dynamic changes in the number of OLs surrounding the hematoma after ICH, we labeled and tracked apoptotic, proliferating, and mature OLs using immunofluorescence staining. ResultsSephin1 treatment improved long-term neurological function after ICH, which was accompanied by a significant alleviation of WMI in the perihematomal region. Our data indicated that Sephin1 dramatically increased the population of OLs in the perihematomal region after ICH by inhibiting OL apoptosis and promoting OL proliferation. Moreover, Sephin1 treatment attenuated neuroinflammation after ICH by inhibiting microglial polarization to the M1 phenotype. In vitro, a co-culture model of primary OLs and microglia demonstrated that Sephin1 preserved the viability of OLs under pro-inflammatory conditions. ConclusionsOur observations suggest that Sephin1 is a promising therapeutic drug to preserve the OLs and alleviate WMI around the hematoma in ICH, highlighting its translational potential to improve long-term neurological recovery in hemorrhagic stroke.

neuroscience↗

bioRxiv2025

Population-specific brain charts reveal Chinese-Western differences in neurodevelopmental trajectories

Human brain charts provide unprecedented opportunities for decoding neurodevelopmental milestones and establishing clinical benchmarks for precision brain medicine 1-7. However, current lifespan brain charts are primarily derived from European and North American cohorts, with Asian populations severely underrepresented. Here, we present the first population-specific brain charts for China, developed through the Chinese Lifespan Brain Mapping Consortium (Phase I) using neuroimaging data from 43,037 participants (aged 0-100 years) across 384 sites nationwide. We establish the lifespan normative trajectories for 296 structural brain phenotypes, encompassing global, subcortical, and cortical measures. Cross-population comparisons with Western brain charts (based on data from 56,339 participants aged 0-100 years) reveal distinct neurodevelopmental patterns in the Chinese population, including prolonged cortical and subcortical maturation, accelerated cerebellar growth, and earlier development of sensorimotor regions relative to paralimbic regions. Crucially, these Chinese-specific charts outperform Western-derived models in predicting healthy brain phenotypes and detecting pathological deviations in Chinese clinical cohorts. These findings highlight the urgent need for diverse, population-representative brain charts to advance equitable precision neuroscience and improve clinical validity across populations.

neuroscience↗

bioRxiv2025

Quantifying Protein-Protein Interaction with a Spatial Attention Kinetic Graph Neural Network

Accurate prioritisation of near-native protein-protein interaction (PPI) models remains a major bot-tleneck in structural biology. Here we present SAKE-PP, a physics-inspired, spatial-attention equivariant graph neural network that directly regresses interface RMSD (iRMSD) without any native references. Trained on docking decoys generated through our novel hierarchical sampling strategy applied to PDB-Bind dataset, SAKE-PP combines force-field-like attention with Laplacian-eigenvector orientation to couple local interaction forces with global topology. On the 2024PDB benchmark comprising 176 het-erodimers, SAKE-PP demonstrates effective optimization and selection of AF3 decoys, achieving improvements of 13.75% based on iRMSD statistics and 12.5% based on DockQ scores. It consistently outperforms the AF3 ranking score in multiple metrics, including overlap, hit rate, and correlation. In zero-shot evaluation of 139 antibody-antigen complexes, SAKE-PP improves the score-iRMSD correlation by 0.4. By unifying geometric deep learning with physics-based realism, SAKE-PP provides a robust, plug-and-play scoring function that streamlines reliable PPI evaluation and accelerates downstream structure-guided drug-design workflows.

bioinformatics↗

bioRxiv2025

GeneChat: A Multi-Modal Large Language Model for Gene Function Prediction

Accurately predicting gene function from DNA sequences remains a fundamental challenge in genomics, particularly given the limited experimental annotation available for most genes. Existing computational approaches often formulate function prediction as a classification task over predefined categories, limiting their flexibility and expressiveness. We introduce GeneChat, a multi-modal large language model designed to generate free-form, natural language descriptions of gene functions directly from nucleotide sequences and textual prompts. GeneChat integrates three components: a DNABERT-2-based gene encoder optimized for long-range genomic context, an adaptor that aligns gene representations with the input space of a large language model, and Vicuna-13B, a fine-tuned LLaMA-2 variant used to produce coherent functional narratives. Trained on over 50,000 genes from the NCBI database, GeneChat outperforms GPT-4o on BLEU and METEOR metrics, demonstrating superior ability to generate accurate, context-aware, and semantically rich descriptions. This work highlights the potential of foundation models for advancing interpretable and scalable gene function prediction in a free-form, language-driven paradigm.

bioinformatics↗

bioRxiv2025

Impact of Peptide Initial Configuration and Membrane Composition on Melittin's Pore-Forming Ability under Unbiased All-Atom Molecular Dynamics Simulations

The rising challenge of antimicrobial resistance has accelerated the search for alternative therapeutics. Antimicrobial peptides (AMPs), a class of naturally occurring defense molecules found across diverse species, are promising candidates. Despite their potent membrane-disrupting activity, the atomic details of the pore formation process remain insufficiently understood. In this study, we employed all-atom molecular dynamics (MD) simulations to investigate the pore formation process of melittin under different initial configurations. Simulations were conducted using three different membrane systems: a pure POPC bilayer, a mammalian membrane model (DOPC:Cholesterol, 9:1), and a bacterial membrane model (DOPE:DOPG, 3:1). For each system, we examined four different starting configurations, in which six melittin peptides were arranged in a star-like pattern. Our results demonstrated that the pore formation process strongly depends on the initial peptide configuration. In one specific initial arrangement (termed as Conf.I), pore formation consistently occurred within 100 nanoseconds, regardless of membrane composition. Furthermore, the simulations revealed that pore formation was more challenging in the mammalian membrane model and even more so in the bacterial membrane model, in comparison with the pure POPC bilayer. These findings are in line with previously reported minimum inhibitory concentration (MIC) and the 50% hemolysis concentration (HC50) of melittin in the literature. Additionally, we identified lysine-7 (K7) as the key residue in determining whether a stable pore can form. In configurations where the K7 side chain formed electrostatic interactions with the phosphate group of a lipid, melittin were anchored to the membrane surface, thereby preventing pore formation. In contrast, simulations of melittin mutants K7A and K7Q showed no such anchoring effect, and thus pore formation was possible in multiple initial configurations. Notably, the K7Q mutation showed a preference for pore formation in bacterial membranes over mammalian membranes, suggesting that reducing toxicity while maintaining antimicrobial efficacy is possible.

biochemistry↗

bioRxiv2025

Interpreting biochemical text with language models:a machine learning framework for reaction extraction and cheminformatic validation

Recent advancements in large language models (LLMs) offer new opportunities for automating the manual curation of biochemical reaction databases from scientific literature. In this study, we present an integrated pipeline that enhances LLM-based extraction of enzymatic reactions with machine learning and cheminformatics-informed validation. Using BRENDA-linked PubMed articles, we evaluate GPT-4s ability to extract reactions and infer missing chemical entities in textual descriptions of enzymatic reactions. Extracted reactions are converted to SMILES and InChI notations before being encoded into molecular fingerprint similarity scores and atom mapping metrics. These cheminformatics metrics are then used to train machine learning classifiers that validate GPT extractions. We employ a Positive-Unlabeled learning approach with synthetic invalid reactions to train various classifiers and assess model performances. The best classifier is then benchmarked on GPT extractions. Our findings show that GPT can accurately infer incomplete reactions and cheminformatics tools can serve as effective predictors of reaction validity. This work demonstrates a scalable framework for automated and reliable curation of enzymatic reaction databases, highlighting the potential of combining LLMs with cheminformatics and machine learning for reliable scientific knowledge extraction. Author SummaryCurating databases of biochemical reactions is a time-consuming and manual task, yet it plays a vital role in advancing research in biology and chemistry. Many scientific articles describe important enzymatic reactions, but often do so in incomplete ways--such as mentioning only the starting molecule or the enzyme, and leaving out the rest. In this work, we explore how recent advancements in artificial intelligence, specifically large language models like GPT, can help extract such information automatically from scientific literature. We show that these models can not only find reactions in text, but also infer missing parts of reactions based on the surrounding context. To make sure these inferred reactions are chemically plausible, we use computational chemistry tools that analyze the structure of the molecules involved. We then train a machine learning model to help us automatically detect which reactions are likely to be valid. This combination of tools offers a new way to speed up and improve how biochemical knowledge is extracted from the growing body of scientific literature. Our study suggests that this kind of automation could help scientists keep biological databases up to date and reduce the burden of manual data entry.

bioinformatics↗