Biology subjects

Li, W.

Publications and source records attributed to Li, W..

At least 145 records · Page 8Linked to original sources

bioRxiv2026

Towards building a World Model to simulate perturbation-induced cellular dynamics by AlphaCell

Predicting cellular responses to perturbations is crucial for therapeutic discovery, yet experimental screening is severely constrained by the combinatorial vastness of biological space. While computational simulations offer a scalable alternative, current models are limited by the incomplete latent representation----mainly relying on highly variable genes in feature representation; the poor genome-wide reconstruction fidelity; and the ungeneralizable dynamic laws across diverse contexts. Consequently, they fail to mechanistically transfer learned dynamics to unseen cellular contexts. To address these systemic flaws, we introduce AlphaCell, a generative Virtual Cell World Model that unifies genome-wise representation with continuous state transition modeling. AlphaCell achieves three synergistic innovations: (1) Latent Manifold Rectification, processing the full protein-coding transcriptome to construct a differentiable Virtual Cell Space, effectively filtering noise while preserving intrinsic cellular topology; (2) Biological Reality Reconstruction, utilizing a massive, knowledge-rich decoder to translate abstract latent states back into high-fidelity, genome-wide expression profiles; and (3) Universal State Transition, applying Optimal Transport Conditional Flow Matching to model perturbations as continuous, deterministic vector fields. By abstracting perturbation mechanisms into generalized dynamic laws, AlphaCell makes robust prediction of perturbation responses in a compositional generalization scenario and enables zero-shot prediction of cellular dynamics in entirely unseen cellular contexts, providing a foundational engine for cellular-context-generalizable perturbation prediction and perturbation-induced cellular dynamics simulation.

bioinformatics↗

bioRxiv2026

Toward Computationally Complete Spatial Omics

Multimodal spatial omics has transformed biology by mapping molecular complexity within intact tissues, yet current technologies remain limited in the number of modalities measured simultaneously and often produce lower-quality data than single-modality assays. We present COSIE, a computational framework that generates high-resolution, multilayered molecular landscapes across tissue sections, individuals, and platforms. COSIE integrates histology, epigenome, transcriptome, proteome, and metabolome into a unified representation. Applied to 12 datasets spanning 10 spatial technologies, eight modalities, and nine tissue types, ranging from thousands of spots to millions of cells, COSIE outperforms existing methods. It resolves tissue structures, enhances noisy measurements, predicts unmeasured modalities, and captures dynamic processes. In human tumors, COSIE identifies invasive subregions linked to clinical outcomes and predicts spatial gene expression in TCGA samples using only histology images. By transforming fragmented data into comprehensive spatial maps, COSIE advances computationally complete spatial omics and the creation of digital tissue twins for biomedicine.

genomics↗

bioRxiv2026

Tissue- and Age-Specific Transcriptomic and Metabolomic Analysis Reveals Regulatory Mechanisms of Ginsenoside Biosynthesis in Panax notoginseng

Ginsenosides are the primary bioactive compounds in Panax notoginseng, but the transcriptional mechanisms governing their tissue- and age-dependent accumulation remain elusive. Here, we integrated targeted metabolomics with transcriptome profiling across four tissues (root, stem, leaf, and flower) and three developmental stages (1-3 years) to investigate the spatiotemporal regulation of ginsenoside biosynthesis. We observed distinct tissue- and age-specific accumulation patterns: roots exhibited a progressive increase in total ginsenoside content during the second and third years, while flowers preferentially accumulated rare protopanaxadiol-type ginsenosides such as Rg3-2. Transcriptomic analysis revealed extensive differential gene expression across tissues and stages, particularly in roots during late development. Clustering and transcription factor (TF) enrichment analyses identified multiple tissue-associated regulatory modules. Four TFs--AT3G12130, SPL9, MYB33, and SPL1--emerged as core candidates based on coordinated expression, promoter motif enrichment, and functional annotation of predicted target genes. Motif analysis further linked these TFs to key biosynthetic genes involved in triterpenoid oxidation and glycosylation, including CYP716A53v2 and CYP716A47. Together, these findings suggest that tissue-specific ginsenoside accumulation in P. notoginseng is associated with coordinated transcriptional regulation of biosynthetic enzymes. This study provides a transcriptomic framework for understanding spatial regulation in ginsenoside biosynthesis and identifies candidate regulators for future functional validation and metabolic engineering.

plant biology↗

bioRxiv2026

Myristoylation licenses disordered viral VP4 protein to anchor to and perforate the membrane through phase separation

The VP4 protein of enteroviruses, such as Coxsackievirus B3, is a small, intrinsically disordered protein essential for perforating the host cell membrane during viral entry. A key feature of VP4 is its N-terminal myristoylation, which is required for infectivity in some enteroviruses but dispensable in others, suggesting a complex and context-dependent role that is not fully understood. The precise biophysical mechanisms by which this lipid anchor enables a disordered protein to breach a membrane remain unresolved. Here, using Coxsackievirus B3 VP4 as a model system and integrating multi-scale molecular dynamics simulations with confocal microscopy, we demonstrate that myristoylation is not a simple membrane tether but a multi-functional regulator that orchestrates VP4 activity through distinct, hierarchical roles. First, it provides the necessary hydrophobic anchor to recruit the disordered VP4 to the membrane interface. Second, the myristoyl group acts as a key molecular driver that promotes the liquid-liquid phase separation of VP4, leading to the formation of dynamic condensates on the membrane surface. These condensates actively remodel the membrane, generating substantial curvature that, in turn, lowers the free energy barrier for VP4 penetration. Furthermore, we find evidence that the myristoyl group plays a third role in stabilizing the final transmembrane pore. Our findings establish a novel paradigm where a single lipid modification empowers a disordered viral protein to form a functionally potent condensate that mechanically primes and physically breaches the target membrane, a mechanism that may explain the conditional myristoylation requirement across enteroviruses.

biophysics↗

bioRxiv2026

Evolutionary Divergence of Structure-Function Coupling between Human and Macaque: Spatial Patterns and Transcriptomic Basis

Structural-functional coupling (SFC) provides critical insights into how the brains structural architecture constrains functional organization for supporting higher-order cognition. Investigating evolutionary differences in SFC between humans and macaques may provide new insights into the neural basis of unique human cognitive abilities. In this study, we analyzed multimodal magnetic resonance imaging data from anesthetized and awake adult rhesus macaques and adult humans to examine cross-species divergences in SFC. Moreover, we integrated transcriptomic data to elucidate the molecular mechanisms underlying evolutionary differences in SFC patterns. We find that humans and macaques exhibit distinct SFC patterns: the human brain shows high SFC in the lateral and medial prefrontal cortex, whereas macaques show high SFC in the sensorimotor cortex. Notably, language-related regions in the human lateral temporal cortex exhibit relatively low SFC. Furthermore, the human whole-brain SFC pattern and the evolutionary differences in SFC between humans and macaques are negatively correlated with cortical evolutionary expansion. By integrating human and macaque transcriptomes, we reveal that the macaque SFC specifically associated genes are primarily involved in basic physiological functions, whereas the human SFC specifically associated genes exhibit evolutionary adaptations in synaptic function, neurotransmitter secretion, and other molecular processes. Moreover, the human-specific genes showing significant overlap with Human Accelerated Regions genes were mainly enriched in cell types of astrocyte and oligodendrocyte and in diseases of schizophrenia and Alzheimers diseases. Overall, these findings advance our understanding of the intricate relationships of SFC in human and macaque brains and provide novel insights in understanding evolutionary conservation and species specificity in cognitive function and gene regulation.

evolutionary biology↗

bioRxiv2026

Immune Checkpoint B7-H3 Inhibition Reprograms Acute Neuroinflammation and Protects the Brain After Acute Ischemic Stroke

BACKGROUND AND PURPOSEThe role of immune checkpoint B7-H3 in acute ischemic stroke prognosis and post-stroke immunosuppression remains uninvestigated, despite the clinical significance of immune checkpoints with inflammaging and post-stroke infections. We recently reported the neuroprotective effects of acute B7-H3 inhibition after stroke in adult and aged mice. In this study, we investigated the mechnsistic association of regulating the cerebral induction of B7-H3 after acute ischemic stroke on brain damage, neuroinflammation, vascular integrity, host defense gene regulation, and functional outcomes. METHODSC57BL/6 mice were subjected to transient middle cerebral artery occlusion and injected (i.v.) with either B7-H3 siRNA or a negative (non-targeting) siRNA at 5 min after reperfusion. At 24 hours of reperfusion, magnetic resonance imaging (MRI) of the mouse brain was performed using a 9.4 T scanner to assess brain damage (T2, ADC, and Kurtosis). Real-time qPCR and NanoString nCounter(R) Neuroinflammation Panels were used to determine the acute changes in overall neuroinflammatory functions that are mediated by B7-H3. Motor function (beam walk, rotarod tests, and grip strength) was assessed between days 1 and 7 of reperfusion. RESULTSEarly inhibition of B7-H3 after stroke significantly reduced the brain damage and promoted the functional outcomes. Post-stroke neuroinflammation was reprogrammed with B7-H3 inhibition towards balancing of neuroprotective anti-inflammatory mechanisms without compromising the immune response that is crucial for preventing post-stroke infections. CONCLUSIONSOur results demonstrate that the induction of B7-H3 during the acute period after stroke is a mediator of post-stroke neuroinflammation and secondary brain damage.

neuroscience↗

bioRxiv2026

Universal physical principles govern the deterministic genesis of protein structure

Protein structure is usually treated as the endpoint of sequence by most AI models, yet its true biological emergence is a process of ordered change, whose logic remains hidden in opaque black box. ProtGenesis creates a bidirectional mirror world: it maps amino acid assembly, elongation and mutation in biological space into quantitative trajectories and ensembles in protein language model representation space, then translates spatial geometry into testable biophysical hypotheses. Across peptides, reporter proteins and protein families, three universal general principles emerged: hierarchical directional assembly (Principle I), quantitative and reproducible structural-emergence trajectories (Principle II) and discrete topological transitions (Principle III) from short- to long-range order. Three novel metrics of spatial features [D,{rho} ,{delta} ], complemented by squared Gaussian Wasserstein-2 measures, located structural anchors, sensitive regions and state transitions, enabling split-protein engineering and programmable protein design. ProtGenesis makes latent representations mechanistically interpretable, offering AI a route to discovering, rather than merely predicting, scientific principles. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=173 HEIGHT=200 SRC="FIGDIR/small/706798v2_ufig1.gif" ALT="Figure 1"> View larger version (60K): org.highwire.dtl.DTLVardef@18f690org.highwire.dtl.DTLVardef@e36903org.highwire.dtl.DTLVardef@35f15org.highwire.dtl.DTLVardef@1577b81_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2026

Event2Flow: Scalable imaging of peripheral and cerebral hemodynamics with event-based vision sensors

Accurate blood flow mapping over mesoscale fields of view is essential for understanding physiological and pathological processes, yet conventional optical methods often rely on bulky high-speed cameras that generate massive datasets with excessive computation burden. Here, we introduce Event2Flow, a compact and data-efficient framework leveraging event-based vision sensors, which asynchronously capture brightness changes with sub-millisecond latency and minimal data redundancy. Event2Flow supports multiple contrast mechanisms for flow measurement, including speckle fluctuation and particle tracking. By correlating the event count with flow velocity through simulations and experiments, we first demonstrate its application in laser speckle imaging for noninvasive mapping of mouse ear vasculature and ethanol-induced hemodynamic changes. When integrated with widefield fluorescence localization microscopy and point spread function engineering, Event2Flow further enables kilohertz-rate particle tracking for rapid 3D velocity quantification in transcranial brain imaging and snapshot flow direction estimations using event polarity. Overall, Event2Flow offers a scalable alternative to conventional high-speed imaging systems for vascular and neuroimaging applications.

neuroscience↗

bioRxiv2026

A Discrete Language of Protein Words for Functional Discovery and Design

Proteins can preserve conserved functions despite extensive sequence and structural divergence, suggesting that functional organization is governed by distributed constraints not captured by conventional representations. Here we develop a hierarchical sequence-based representation framework that compresses proteins into context-dependent latent states while preserving multiscale organizational information. Using this framework, we identified previously uncharacterized ciliary proteins lacking detectable sequence and structure homology, including ADMAP1, which is required for normal sperm axonemal organization and motility in mice. Discrete latent protein states captured species-level organizational signatures correlated with major evolutionary groups and revealed expansion of intrinsically disordered regulatory environments in eukaryotes. Autoregressive sampling within this latent space further enabled design of synthetic actin-remodeling proteins that maintained robust F-actin severing activity despite extensive sequence rewiring across key functional interfaces. These findings demonstrate that distributed protein organization can be inferred directly from sequence, linking functional discovery, evolutionary analysis, and protein design within a shared representational framework.

bioinformatics↗

bioRxiv2026

Cigarette smoke induces colon cancer by regulating the gut microbiota and related metabolites

The causal relationship between smoking and colorectal cancer (CRC) remains unclear. In this study, a cigarette smoke-exposed mouse model demonstrated that smoking significantly increased CRC incidence by inducing gut microbiota dysbiosis and altering related metabolites. Smoke exposure reduced beneficial bacteria (e.g., Lactobacillus), increased harmful bacteria (e.g., Firmicutes and Clostridium), elevated metabolites such as histamine, and suppressed the tumor suppressor genes PARG, CPT2, and ALDH1A1, thereby promoting tumor development. Functional assays in CRC cell lines further confirmed that CPT2 knockdown enhanced malignant phenotypes, including proliferation, migration, and invasion. Clinical analysis showed that these genes were markedly downregulated in smoking-related CRC patients, with strong diagnostic value (AUC > 0.8). ConclusionSmoking promotes CRC by inducing gut microbiota dysbiosis, metabolic reprogramming, and suppression of tumor suppressor genes, particularly CPT2. These findings highlight the importance of smoking cessation in CRC prevention and provide potential biomarkers for early diagnosis and therapeutic intervention.

bioinformatics↗

bioRxiv2026

TLR5 drives metabolic dysfunction-associated steatohepatitis through lipid- and flagellin-induced hepatocyte injury signalling

Liver fibrosis is a strong predictor of clinical outcomes in metabolic dysfunction-associated steatohepatitis (MASH). Fibrosis is a consequence of persistent liver cell injury and inflammation in which Toll-like receptors (TLRs) play a key initiating role. Here we test the hypothesis that TLR5 is involved in the development of MASH and fibrosis using a combination of clinical data from multiple independent patient cohorts, single cell liver transcriptomics and human in vitro and ex vivo models. Hepatic TLR5 expression, but not TLR2 or TLR4, is associated with liver fibrosis and mortality. Plasma levels of TLR5s cognate ligand flagellin are increased in MASH with advanced fibrosis and fall with liver disease improvement. Mechanistically, we identify two parallel TLR5-mediated routes to hepatocyte injury: one elicited by flagellin and the other indirectly by lipid injury. Furthermore, hepatocyte TLR5 inhibition abrogates paracrine activation of hepatic stellate cells to suppress collagen production. This is also seen ex vivo in patient-derived precision-cut liver slices where TLR5 inhibition significantly reduces lipid-induced collagen deposition. These findings reveal a new role for TLR5 signalling, specifically in the development of advanced MASH fibrosis and may offer a novel disease-specific therapeutic approach.

molecular biology↗

bioRxiv2026

The BELL-type homeobox transcription factor RLC3/OsBLH4 controls leaf rolling and drought tolerance via KNOX-BELL-lignin regulatory network in rice

O_LIModerate leaf rolling in rice is crucial for plant architecture and stress adaptation, but its molecular regulation remains unclear. We investigated the role of RLC3/OsBLH4, a BELL-type homeobox transcription factor, in controlling leaf rolling and drought tolerance, addressing gaps in lignin biosynthesis and cell wall development mechanisms. C_LIO_LIWe used gene map-based cloning (rlc3-1, rlc3-2), CRISPR/Cas9 knockout lines (rlc3-ko#11, rlc3-ko#12), and allelic complementation to validate RLC3s function. Additionally, we employed biochemical assays, gene expression analysis, and protein interaction studies to explore its regulatory network. C_LIO_LIRLC3 mutations impaired lignin biosynthesis and secondary cell wall formation, reducing bulliform cells area and causing midrib defects. These structural abnormalities accelerated water loss, leading to excessive leaf rolling and compromised drought tolerance. Mechanistically, RLC3 directly activates lignin synthesis genes (OsPAL5, OsCOMT5, OsCCR4, OsCAld5H1) and interacts with KNOX transcription factors (OSH1, OSH45, OSH71) to form a KNOX-BELL complex, further regulating lignin content and cell wall development. C_LIO_LIRLC3 orchestrates lignin deposition and secondary cell wall development to control leaf rolling, water transport, and drought tolerance. This study reveals a novel KNOX-BELL-lignin regulatory module governing leaf morphology and stress adaptation, offering targets for crop improvement under drought conditions. C_LI

plant biology↗

bioRxiv2026

Unravelling genome-wide mosaic microsatellite mutations at single-cell resolution

Short tandem repeats (STRs), or microsatellites, are highly mutable genomic elements that modulate gene regulations and are implicated in a range of human diseases. However, detecting mosaic STR mutations at single-cell resolution remains challenging due to both technical and biological complexities. To address this, we developed BayesMonSTR, a robust algorithm that enables accurate detection of mosaic STR mutations. Using this tool in single-cell analysis of human tissues, we reveal an accumulation of longer mosaic STR insertions and deletions (indels) in aging mitotic and post-mitotic cells. Strikingly, prefrontal cortex (PFC) neurons accumulate a higher burden of STR mutations than B cells or lung epithelium, with aged neurons exhibiting a particularly pronounced increase in longer STR deletions. These mutations are enriched at transcription start sites (TSSs) and active enhancers of highly expressed genes. Our work establishes a foundation for genome-wide, hypothesis-free discovery of disease-associated mosaic STR mutations and reveals a previously unexplored landscape of mosaic STR variation in development and aging.

bioinformatics↗

bioRxiv2026

Cross-species graph-embedding unmasks the ageing microenvironment as a key determinant of pancreatic cancer malignant cell biology and therapy response

Pancreatic ductal adenocarcinoma (PDAC) has a dismal prognosis and is characterised by an extensive pro-tumorigenic stroma. Although most PDAC cases occur in older patients, the impact of ageing on malignant-stromal interactions and therapy response remains poorly understood. Here, we established orthotopically-grafted organoid-derived PDAC models across three murine age groups to characterise changes in the PDAC stroma and malignant cells with ageing. Cross-species analyses of tumour transcriptomes using a graph-embedding approach showed that integrating mouse models of different ages better captures the diversity of human PDAC, and that aged models more faithfully recapitulate the biology of older patients with PDAC. We also demonstrated that aged PDAC models have a more inflammatory stroma than that of younger tumours, shaping the malignant cell transcriptome. Finally, graph-embedding identified IRAK4 as a candidate therapeutic vulnerability in aged, but not young, KRAS- and p53-mutant PDAC, which we validated in preclinical drug studies. These findings highlight how ageing is a critical determinant of PDAC biology and associated therapeutic vulnerabilities, which should be an important consideration when designing disease models for preclinical development of precision therapies.

cancer biology↗

bioRxiv2026

HistoSweep enables cellular-resolution tissue quality control for gigapixel images in digital pathology and spatial omics

High-resolution histology images are indispensable for pathology and increasingly serve as the structural backbone for spatial omics. Yet whole-slide images (WSIs) frequently contain artifacts, acellular voids, and background regions that, when included in computational workflows, introduce noise, degrade model accuracy, and compromise biological interpretation. Existing tools provide only coarse foreground-background separation, leaving a gap in fine-grained quality control (QC). Here we present HistoSweep, a scalable framework that generates morphology-aware tissue masks at cellular resolution. By integrating density filtering, texture descriptors, and adaptive thresholding, HistoSweep systematically removes non-informative tissue regions while preserving biologically meaningful microstructures. It processes billion-pixel WSIs in minutes on standard CPUs, requiring no GPU acceleration, and is deployable across research and clinical settings. Across 25 WSIs spanning distinct tissues, disease states, and spatial omics platforms, HistoSweep consistently outperformed existing methods. It enhanced visualization and segmentation, improved virtual cell type predictions, and safeguarded spatial transcriptomics integrity by detecting transcript leakage and transcript-histology misalignment. By enabling fine-grained, scalable QC, HistoSweep provides a foundational preprocessing step for reliable and reproducible digital pathology and spatial omics analyses.

bioinformatics↗

bioRxiv2026

Macrophage Antigen Presentation Is Unleashed by Pan-RAS Inhibition to Promote Antitumor Immunity

Pan-RAS inhibitors have emerged as potent targeted therapies designed to suppress oncogenic signaling in RAS-mutant tumors. However, whether these inhibitors possess therapeutic utility beyond tumors with intrinsic RAS dependency remains unclear. Here we show that the clinical-stage pan-RAS(ON) inhibitor RMC-6236 (Daraxonrasib) induces robust regression of NRAS-wild-type B16 tumors in immunocompetent mice, a response mediated by a mechanism independent of direct tumor-cell targeting. Instead, RMC-6236 acts by remodeling the tumor microenvironment, specifically enhancing the antigen-presenting capacity of tumor-associated macrophages through upregulated MHC-I expression. This anti-tumor efficacy is abrogated in T cell-deficient hosts, identifying a macrophage-dependent, T cell-mediated mechanism of control. Mechanistically, RMC-6236 does not directly block canonical RAS signaling in macrophages. Rather, it suppresses a Myc-Hdac2 epigenetic program, leading to enhanced histone H4 acetylation at the Nlrc5 locus and subsequent activation of the MHC-I master regulator NLRC5. Together, our findings reveal a macrophage-centered, epigenetically wired immune mechanism for pan-RAS inhibition that broadens the therapeutic scope of targeting the "undruggable" RAS pathway beyond tumor genotype towards systemic immune modulation.

immunology↗

bioRxiv2026

Identification a Compact Promoter using a New Promoter Selection Strategy and Engineering Hybrid Pol II/III Enable Efficient Genome Editing in Human Retinal Ganglion Cells

Promoters and vectors are critical components of gene therapy, enabling the delivery and expression of therapeutic genes to correct both loss- and gain-of-function mutations. Adeno-associated virus (AAV) vectors are the leading platform for in vivo gene delivery; however, the widely used Streptococcus pyogenes Cas9 (SpCas9, 4.1 kb) approaches the AAV packaging limit of 4.7 kb. This constraint often necessitates dual-vector systems, which reduce therapeutic efficiency, or the use of smaller nucleases such as SaCas9 (3.2 kb) and AacCas12b (3.4 kb), which have lower PAM site frequencies. To enhance promoter selection for gene therapy applications, we developed a strategy to identify compact, cell-preferred RNA polymerase II (Pol II) promoters. Analysis of approximately 300 compact Pol II promoters revealed that exogenous expression levels in one cell type correlate more strongly with those in other cell types than with endogenous expression, underscoring the importance of exogenous expression efficiency in promoter selection. Using this approach, we identified a compact Pol II promoter #2 (Pro2, 133 bp) that drives robust transgene expression in human retinal ganglion cells (RGCs). To enable single-AAV delivery of SpCas9, we analyzed three commonly used Pol III promoters (H1, 7SK and U6) and determined their minimal functional lengths using a CRISPR/Cas9 reporter assay. We further engineered three compact hybrid Pol II/III promoters which combined pro2 with minimal H1, 7SK and U6 (276, 294, and 323 bp) capable of co-expressing SpCas9 and gRNA, enabling efficient genome editing in both transfected HEK293 cells (approaching 100%) and human RGCs (up to 55.9%) from human stem cell-derived retinal ganglion cells (RGCs). Together, these findings establish a framework for developing single-AAV CRISPR-based gene therapy strategies. Authors contributionsPWZ and DJZ conceived the study, designed the experiments, performed data analysis and interpretation, and were the primary contributors to manuscript writing. STZ played a key role in data collection and correlation analysis. YYC, SL, LF, CJK, YD, CAB, JC, and DW contributed to the execution of essential experiments and subsequent data analysis. All authors have read and approved the final manuscript. Declaration of interestsThe authors declare no conflicts of interest.

molecular biology↗

bioRxiv2026

Dual-channel graph learning reveals similarity and complementarity in protein-protein interaction networks

Protein-protein interactions (PPIs) are governed by two fundamental interfacial mechanisms: similarity-driven, often involving symmetric structural motifs, and complementarity-driven, arising from geometric and physicochemical matching between binding surfaces. Despite their biological significance, computational models have largely overlooked the coexistence and interplay of these twofold interaction modes. Here, we introduce DMG-PPI, a dual-channel graph neural network framework that jointly models similarity and complementarity in PPI networks, extending prior heterophilous GNN concepts to explicitly disentangle these dual interaction modes. The model consists of two parallel message-passing pathways: Alignment Message Passing (AMP), which aggregates signals from similar neighbors to capture symmetric interfaces, and Divergence Message Passing (DMP), which contrasts node features to extract complementary binding patterns. The signals captured by AMP and DMP are integrated via an adaptive fusion strategy within each block, and the outputs of blocks are aggregated using the MixHop framework to encode higher-order interaction patterns. DMG-PPI substantially outperforms state-of-the-art methods on classical benchmark datasets, achieving a 7.19% improvement in Micro-F1 over the second-best method. Additionally, the dual-channel framework provides interpretable insights into key binding residues by identifying interfacial mechanisms. Overall, DMG-PPI serves as a powerful tool that reveals the mechanisms behind accurate PPI predictions and facilitates downstream biological analysis. Author summaryProteins interact through multiple interfacial principles rather than a single uniform mechanism, reflecting diverse modes of molecular recognition between binding partners. Accurately modeling this heterogeneity remains a central challenge in protein-protein interaction (PPI) prediction. Existing computational approaches often overlook such complexity and treat all interactions as arising from a dominant pattern. To address this limitation, we present DMG-PPI, a dual-channel graph neural network framework designed to account for distinct interaction principles within a unified model. By modeling both similarity-based and complementarity-based interaction signals, DMG-PPI enables a more comprehensive representation of interaction patterns in PPI networks. Evaluations on widely used benchmark datasets demonstrate that our approach achieves robust and consistent improvements over existing methods. Beyond prediction accuracy, DMG-PPI provides biologically meaningful interpretability by highlighting key residues and revealing distinct interfacial interaction mechanisms, offering valuable insights for downstream structural and functional studies.

bioinformatics↗