Biology subjects

Hao, M.

Publications and source records attributed to Hao, M..

At least 19 recordsLinked to original sources

bioRxiv2026

Structural Modeling Identifies a Conserved ADAR1 dsRBDIII Motif Predicted to Interact with PKR at Thr451

Protein kinase R (PKR) is a key cellular dsRNA sensor whose activation is regulated by ADAR1. Although ADAR1 PKR association is established, its molecular determinants remain poorly understood. Molecular docking identified a conserved W768/F769/P770 (WFP) motif within ADAR1 dsRBD III as a potential PKR interaction interface. Disruption of this motif altered the predicted binding configuration. However, coimmunoprecipitation revealed no significant reduction in PKR association following individual substitutions or deletion of the WFP motif, although the deletion showed a consistent trend toward reduced binding. RNase treatment reduced PKR coimmunoprecipitation approximately tenfold without abolishing the interaction. These findings suggest that ADAR1 PKR association is predominantly RNA dependent, potentially involving shared dsRNA substrates that facilitate direct protein protein contacts, with the WFP motif contributing to interaction stability.

molecular biology↗

bioRxiv2026

DPF3a couples H3K36me2-dependent chromatin remodeling to genome architecture during myogenesis

Chromatin remodelers are generally thought to regulate gene expression through nucleosome mobilization and modulation of local chromatin accessibility. Whether these complexes can instead control transcription through higher-order chromatin organization remains poorly understood. Here we identify the BAF subunit DPF3a as an essential regulator of skeletal muscle regeneration and myogenic differentiation. DPF3a functions through association with the H3K36me2/3 reader HRP2 and preferentially localizes to a distinctive chromatin state characterized by focal depressions within broad H3K36me2 domains. Biochemical reconstitution demonstrates that H3K36 methylation directly enhances remodeling activity of the DPF3a-containing cBAF complex. Unexpectedly, despite profound transcriptional defects, loss of DPF3a produces minimal changes in local chromatin accessibility. Instead, DPF3a is required for long-range chromatin looping associated with activation of myogenic genes. Together, our findings uncover a non-canonical mechanism whereby a chromatin remodeler regulates transcription primarily through three-dimensional genome organization rather than local accessibility control, and establish histone modification-guided chromatin remodeling as a key principle in gene regulation.

molecular biology↗

bioRxiv2026

A Metabolic Enzyme, Pyruvate Carboxylase, Functions as a Sequence-Selective Small RNA Sensor for Antiviral Immunity

Interleukin-27 (IL-27) is an anti-HIV cytokine that induces 14 novel microRNAs (miRNAs) in primary CD4(+) T cells. We previously reported that transfection of two of these miRNA mimics, miRTC10 and miRTC14 into human primary macrophages, differentially induced interferon (IFN)-2, -8, -13, and -{lambda}1 expression. However, the mechanism underlying this activation remains unclear. Here, we demonstrate that miRTC14 does not directly target canonical IFN-regulatory genes but instead engages with pyruvate carboxylase (PC) and laboratory of genetics and physiology-2 (LGP2/DHX58) as direct binding partner proteins. Functional analysis revealed that miRTC14 transfection induces IFN expression by more than 100-fold (p<0.001), whereas PC or LGP2 depletion by siRNAs markedly attenuated this response (50-100-fold reduction, p<0.01). Reconstitution of PC and LGP2 in deficient HEK293 cells restored miRTC14-driven IFN induction. Notably, IFN activation depended on sequence features at the duplex termini. However, PC-miRTC14-LGP2 axis activates TBK1-dependent phosphorylation of IRF3/7, similarly to canonical RNA sensors (RIG-I/MDA5), but this process induced differential IFN subtype. These findings establish PC as a miRNA-binding protein and define a previously unrecognized RNA-sensing mechanism, linking metabolic enzymes to RNA sequence-dependent innate immunity. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=189 SRC="FIGDIR/small/735367v2_ufig1.gif" ALT="Figure 1"> View larger version (42K): org.highwire.dtl.DTLVardef@ea70d3org.highwire.dtl.DTLVardef@1953a73org.highwire.dtl.DTLVardef@305f16org.highwire.dtl.DTLVardef@1b2daff_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOGraphical Abstract:C_FLOATNO microRNA sequence-dependent activation of innate immunity by pyruvate carboxylase (PC). PC-miRTC14-LGP2 axis induces TBK1 phosphorylation via an as-yet-unidentified adapter(s). The activated TBK1 (p-TBK1) subsequently phosphorylates IRF3 and/or IRF7 leading to selective IFN subtypes (IFN-2, IFN-8, IFN-13, and IFN-{lambda}1). (Created in BioRender. Kariyawasam, U. (2027) https://BioRender.com/u82s04p) C_FIG

immunology↗

bioRxiv2026

Multi-omic and spatial analysis reveals tumour-derived paracrine signals drive suppressive macrophage polarisation via activation of the cAMP-CREB axis in glioblastoma

Tumor-associated macrophages (TAMs) are key mediators of tumor immunosuppression, yet the factors governing their polarization remain poorly understood, especially in highly immunosuppressive cancers, including cancers affecting the central nervous system. This study investigates the molecular pathways underlying TAM polarization in glioblastoma, one of the most immunosuppressive cancer types. Using a multi-omics approach integrating spatial proteomics, RNA-sequencing, and proteomic profiling of tumor cells and macrophages, we demonstrate that circulating monocytes polarize toward an immunosuppressive state when they exit tumor blood vessels, in response to glioblastoma cell-secreted cytokines. In situ and in vitro data shows that macrophage polarization is regulated via the cAMP-CREB signaling-transcription axis.

cancer biology↗

bioRxiv2026

Autoregressive forecasting of future single-cell state transitions

Predicting future cell state transitions from static transcriptomic snapshots is vital for understanding and steering cellular dynamics. Although existing trajectory and RNA velocity methods can infer local transitions within observed data, they are fundamentally limited in extrapolating long-range cellular dynamics. Here, we introduce CellTempo, an autoregressive generative framework that enables long-range forecasting of cell state trajectories directly from static transcriptomic snapshots. CellTempo is pretrained on scBaseTraj, a large-scale collection of multi-step cellular transition sequences constructed by integrating diverse experimental technologies and data sources. Across diverse biological systems, CellTempo accurately predicts long-range lineage progression and reconstructs unseen cellular potential landscapes. It further distinguishes transient perturbation responses from long-term cell-fate changes, enabling prioritization of candidate compounds for cell-fate engineering. These results demonstrate that long-range cellular dynamics are inferable from static observations, establishing a foundation for predictive and controllable modeling of cell fate.

bioinformatics↗

bioRxiv2026

How different AI models understand cells differently

AI single-cell foundation models (scFMs) are believed to be able to learn essential relations in cell transcriptomics with the attention modules in Transformer, but there is no method to reveal what they actually learned. We observed that different models may grasp different aspects of relations. To unravel the mystery, we propose scGeneLens, a framework for dissecting how scFMs perceive cells. We employed a sparse block attention to replace the original attention mechanism to concentrate attentions into a few dominant gene-gene relations, used attention propagation to trace how the relations propagate across Transformer layers, and used integrated gradients to disentangle the relative contributions of gene identity and expression in cell representations. We applied it to scFoundation and scGPT and show that they exhibit pluralistic perceptions of cells: scFoundation emphasizes relations among cell-type marker genes, resulting in stronger cell-type separability, whereas scGPT focuses more on genes involved in shared cellular pathways and core biological activities, leading to representations that generalize across conditions. The framework provides a unified lens for probing what scFMs learn about cells and offers actionable insights for the design of future cellular foundation models. Our code can be seen in https://anonymous.4open.science/r/scGeneLens-B771/.

bioinformatics↗

bioRxiv2025

BrainBridge Characterizes Key Factors affecting Alzheimer's Disease and Associated Phenotypes

Single-cell RNA sequencing (scRNA-seq) has significantly advanced our understanding of Alzheimers disease and aging by revealing cellular heterogeneity and shifts in cell-type composition between diseased/old and healthy/young individuals. However, few existing studies utilize the rich information in single-cell transcriptomic atlases for robust patient-level modeling and biological feature selection. To address this gap, we present BrainBridge, a deep learning-based framework designed to integrate atlas-scale single-cell transcriptomic data with phenotypic information to model the biomolecular complexity of the human brain. BrainBridge functions both as a powerful predictor and an embedding model for representing sample-level expression profiles and covariates through comprehensive benchmarking. We also demonstrate its effectiveness in prioritizing key genes and cell types associated with disease progression, aging, and sex differences. We further validate our findings using external resources, including genome-wide and epigenome-wide association studies (GWAS and EWAS), spatial transcriptomics, and perturb-seq experiments. Finally, we deploy BrainBridge within an interactive, agent-powered interface that enables intuitive and user-friendly model interactions, promoting broader accessibility and application in biomedical research.

bioinformatics↗

bioRxiv2025

Emergence of hypervirulent Klebsiella pneumoniae ST111 with a novel virulence plasmid causing pyogenic liver abscess

Hypervirulent Klebsiella pneumoniae (hvKp)-associated pyogenic liver abscess is typically linked to K1/K2 capsule serotypes and canonical lineages such as ST23, ST65, and ST86. Cases outside these lineages are rare but pose significant public health concerns. We investigated an unusual liver abscess in an immunocompetent patient caused by hvKp sequence type 111 (ST111). Five isolates (KP1-KP5) were recovered, all belonging to the same clonal lineage. Genomic analysis revealed a 181 kb IncHI1B/IncFIB virulence plasmid encoding rmpADC and the salmochelin cluster (iroBCDN), but lacking aerobactin (iuc) and rmpA2 gene loci. Notably, this plasmid also carried the yersiniabactin locus (ybt), which is typically located on the chromosome. Comparative genomics indicated that similar ST111 strains have been reported only in China, suggesting clonal expansion of a previously unrecognized hvKp lineage. Functional assays demonstrated that plasmid curing abolished the hypermucoviscous phenotype, reduced capsule production, and completely attenuated virulence in murine models, confirming the plasmid as the major driver of hypervirulence. Interestingly, plasmid loss enhanced biofilm formation and relieved growth burden, highlighting fitness trade-offs associated with virulence. This study documents a rare case of hvKp ST111 liver abscess and identifies a novel plasmid conferring hypervirulence. Our findings expand the known diversity of hvKp and underscore that non-canonical lineages can acquire potent virulence determinants enabling severe disease. Continued clinical and genomic surveillance is essential to detect emerging hvKp clones before widespread dissemination.

microbiology↗

bioRxiv2025

Mechanical Disturbance of Trichomes Enhances Cd2+ Uptake via Malate-Mediated Stomatal Opening in Arabidopsis

Phytoremediation of cadmium-contaminated soils relies on efficient metal uptake and translocation in plants. We previously found that gentle mechanical stimulation of non-glandular trichomes by brushing significantly enhances shoot Cd{superscript 2} accumulation in Arabidopsis thaliana, although the underlying physiological mechanism remained unclear. Here, we demonstrate that trichome brushing specifically induces stomatal opening, elevates transpiration, and promotes Cd{superscript 2} uptake, as evidenced by measurements of stomatal aperture, stomatal conductance, transpiration rate, and Cd{superscript 2} concentration in wild-type and mutant leaves. These responses are absent in the glabrous gl1-1 mutant and are not elicited by brushing epidermal pavement cells, indicating a trichome-specific mechanosensory pathway. Mechanistically, this pathway requires MPK3/MPK6 activity and employs malic acid as a key regulator of stomatal aperture, functioning independently of canonical ABA signaling. Our findings uncover a previously unrecognized trichome-to-stomatal signaling cascade that links trichome mechanostimulation to enhanced heavy metal accumulation.

plant biology↗

bioRxiv2025

Structural and Functional Insights into HBx-Smc6 Targeting for HBV Inhibition

The hepatitis B virus (HBV) X protein (HBx) is a multifunctional regulator essential for HBV replication and HBV-associated hepatocarcinogenesis. Despite its central role, the molecular basis of HBx function has remained elusive. Here, we present the first cryo-electron microscopy structure of the human HBx-CRL4-Smc5/6 complex at 3.1 [A] resolution. In this reconstituted ten-subunit assembly, HBx adopts a Zn2 {square}-stabilized Y-shaped architecture that simultaneously engages the DDB1 and the Smc6 subunit. A composite helix-turn-helix (HTH) pocket in HBx accommodates a conserved "Leucine Key" motif (LRCKL) on Smc6, forming a critical interface essential for complex stability and function. Molecular docking and biochemical validation reveal that the compound Tranilast binds this HTH pocket, disrupts the HBx-Smc6 interaction, and suppresses HBV replication. These findings define the structural mechanism by which HBx counteracts host restriction and establish the HBx-Smc6 interface as a previously unrecognized and druggable target for antiviral intervention.

cell biology↗

bioRxiv2025

pGGTOX: a versatile plasmid platform for recombination-based genome editing across Enterobacteriaceae

Precise genome editing in Enterobacteriaceae is essential for studying gene function, pathogenesis, and antimicrobial resistance, yet many current systems face host-specific and efficiency limitations. We developed pGGTOX, a modular plasmid platform that enables efficient homologous recombination-mediated genome editing across diverse Enterobacteriaceae, including Escherichia coli, Klebsiella pneumoniae, Salmonella enterica, and Enterobacter intestinihominis. The system integrates a rhamnose-inducible toxin (MqsR) for stringent counterselection, a sfGFP reporter for visual tracking of recombination events, Golden Gate cloning for rapid assembly of homologous arms, an FRT-flanked resistance cassette for marker removal, and an oriT sequence for conjugative transfer. Together with the companion plasmid pCP20-oriT, pGGTOX supports precise, marker-free genomic modification. Using pGGTOX, we achieved targeted deletions of dapA in E. coli and mrkCD in carbapenem-resistant K. pneumoniae, both with 100% efficiency. The dapA mutant exhibited diaminopimelate auxotrophy, while mrkCD deletion markedly reduced biofilm formation, consistent with the loss of function associated with these genes. pGGTOX also enabled deletion of a 43.1-kb type IV secretion gene cluster (tra) from an IncN/FII plasmid in E. intestinihominis and insertion of a 10-kb CRISPR-Cas9 plasmid-curing module (pCasCure) into an S. enterica IncX1 plasmid. Deletion of the tra gene cluster resulted in a substantial reduction in plasmid conjugation efficiency. Conjugative transfer of the engineered IncX1-pCasCure plasmid into K. pneumoniae facilitated CRISPR-mediated curing of blaKPC, sensitizing carbapenem resistance to susceptibility. In summary, pGGTOX provides a versatile, efficient, and broadly applicable platform for genome engineering and CRISPR delivery in Enterobacteriaceae, expanding the toolkit for bacterial genetics and translational antimicrobial research. ImportancePrecise genetic manipulation in Enterobacteriaceae remains a major technical challenge, particularly for non-model or multidrug-resistant strains. We developed pGGTOX, a versatile and broadly applicable plasmid platform that enables efficient, marker-free genome editing through homologous recombination. By integrating stringent counterselection, visual screening, modular cloning, and conjugative transfer, pGGTOX simplifies construction and streamlines editing across multiple clinically relevant species. We demonstrate its utility in deleting chromosomal and plasmid-borne loci, inserting large genetic modules, and delivering CRISPR-Cas9 systems for targeted elimination of antibiotic resistance genes. This platform expands the molecular toolkit for functional genomics and provides a powerful new strategy for dissecting bacterial virulence, resistance, and plasmid biology.

microbiology↗

bioRxiv2025

Distinct and Shared Impacts of Virulence Plasmids on the Phenotype and Transcriptome in Convergent Carbapenem-Resistant and Hypervirulent Klebsiella pneumoniae

The global rise of convergent carbapenem-resistant and hypervirulent Klebsiella pneumoniae (CR-hvKp) represents a major clinical challenge, yet the role of virulence plasmids (pVirs) in shaping bacterial physiology and pathogenicity remains incompletely understood. Using a CRISPR-Cas9-based curing system, we precisely eliminated pVirs from two clinical CR-hvKp strains with distinct genetic backgrounds, ST23-KL1 (blaNDM-1) and ST11-KL64 (blaKPC-2). Loss of pVir conferred fitness advantages in vitro, reduced capsule production and hypermucoviscosity, and promoted biofilm formation, while markedly attenuating virulence in murine sepsis models. Despite this reduction, the pVir-cured ST23-KL1 strain retained higher virulence than the pVir-cured ST11-KL64 strain, underscoring the contribution of chromosomal background to pathogenic potential. Transcriptomic profiling revealed both shared and strain-specific transcriptional responses to pVir deletion, with broader perturbations observed in the ST11-KL64 strain. pVir removal had limited effects on antibiotic MICs. Complementation experiments further demonstrated differential regulatory roles of the rmpADC and rmpA2D2 operons in capsule expression and hypermucoviscosity across the two strains. Together, these findings establish pVirs as central determinants of CR-hvKp virulence and highlight complex host-plasmid interactions that influence bacterial adaptation and pathogenicity. IMPORTANCEThe emergence of carbapenem-resistant and hypervirulent Klebsiella pneumoniae (CR-hvKp) poses a critical threat to global health, yet the contribution of virulence plasmids (pVirs) to bacterial fitness and pathogenicity remains poorly defined. By employing a CRISPR-Cas9-based curing strategy, we dissected the role of pVirs in two genetically distinct CR-hvKp strains and uncovered their multifaceted impact on capsule production, hypermucoviscosity, biofilm formation, and virulence. Our findings reveal that pVir loss confers fitness advantages in vitro while attenuating virulence in vivo, with strain-specific transcriptional responses and differential regulation by rmp operons. These results underscore the complex interplay between plasmid-encoded and chromosomal determinants in shaping CR-hvKp pathogenicity and adaptation, offering mechanistic insights that may inform future therapeutic strategies targeting plasmid-mediated virulence.

microbiology↗

bioRxiv2025

Thermogenic Adipose ADH5 Counteracts Age-related Metabolic Decline

Aging-associated decline in brown adipose tissue (BAT) function and mass contributes to energy and metabolic homeostasis disruption. Alcohol dehydrogenase 5 (ADH5) is a major denitrosylase that prevents cellular nitro-thiol redox imbalance, an essential feature of aging. However, the functional significance of BAT ADH5 in the context of aging is largely unknown. Here, we aimed to investigate the role of BAT ADH5 in protecting against age-related metabolic dysfunction. We show that aging elevates BAT protein S-nitrosylation modification and downregulates ADH5 in mice. Furthermore, BAT ADH5-deletion accelerates BAT senescence and aging-associated declines in metabolic homeostasis and cognition. Mechanistically, we found that aging inactivates BAT Adh5 by suppressing heat shock factor 1 (HSF1), a well-recognized proteostasis regulator. Moreover, pharmacologically enhancing HSF1 improved age-related BAT senescence, metabolic decline, and cognitive dysfunction. Together, these findings suggest that the BAT HSF1-ADH5 signaling cascade plays a key role in protecting against age-related systemic functional decline. Ultimately, unraveling the role of thermogenic adipose nitrosative signaling will provide novel insights into the interplay between BAT nitric oxide bioactivity and metabolism in the context of aging. HighlightsO_LIAging elevates general protein S-nitrosylation while downregulating ADH5 expression in BAT. C_LIO_LILoss of thermogenic adipose tissue ADH5 accelerates BAT senescence and exacerbates aging-associated metabolic and cognitive dysfunctions. C_LIO_LIDisruption of HSF1-activated Adh5 in BAT contributes to aging-associated metabolic and cognitive impairments. C_LIO_LIPharmacological targeting of BAT HSF1 alleviates aging-related BAT senescent and systemic declines. C_LI

physiology↗

bioRxiv2025

PerTurboAgent: A Self-Planning Agent for Boosting Sequential Perturb-seq Experiments

Understanding how genetic interventions affect a cells phenotype is key to revealing causal gene regulatory mechanisms and finding novel drug targets. Pooled, high-content perturbation screening methods like Perturb-seq allow us to assess the impact of each of a large number of genetic interventions on a rich cellular profile of RNA or other features, facilitating such discoveries. However, the overall scale of perturbations, especially when considering combinations of genes for perturbation, cannot be tackled exhaustively in the lab. An alternative is to use an iterative design: By leveraging the modularity and sparsity of gene circuits along with prior biological knowledge, we can predict the impact of unseen genetic perturbations on these profiles and group genes with similar effects into co-functional modules, followed by a new round of Perturb-Seq to test these predictions and improve the overall performance of the model. These iterative cycles of experiment and prediction allow prioritizing genes for testing, maximizing the knowledge gleaned from fixed experimental resources, and opening the way to learn general predictive models. Designing these experiments requires a system that can analyze a cellular system, incorporate new and existing knowledge, use statistical tools, predict the effects of unseen perturbations, and prioritize the set of perturbations for the next iteration. These can be time-consuming tasks for scientists and require multiple different skills. Here, we developed PerTurboAgent, an LLM-based agent that excels in predicting candidate gene panels for iterative Perturb-Seq experiments through self-directed data analysis and knowledge retrieval. We evaluated PerTurboAgent based on its ability to identify genes with a phenotypic impact on gene expression upon perturbation in genome-scale perturbation data. PerTurboAgent outperforms existing agent-based and active learning strategies, offering an efficient and understandable approach to designing sequential perturbation experiments.

bioinformatics↗

bioRxiv2025

GeST: Towards Building A Generative Pretrained Transformer for Learning Cellular Spatial Context

Learning spatial context of cells through pretraining on spatial transcriptomics (ST) data may empower us to systematically decipher tissue organization and cellular interactions. Yet, transformer-based generative models often focus on modeling individual cells, neglecting the intricate spatial relationships within them. We develop GeST, a deep transformer model that is pretrained by a novel spatially informed generation task: Predict cellular expression profile of a given location based on the information from its neighboring cells. We propose a spatial attention mechanism for efficient pretraining, a flexible serialization strategy for converting ST data into sequences, and a cell tokenization method for quantizing gene expression profiles. We pretrained GeST on large-scale ST datasets of different ST technologies and demonstrated its superior performance in generating unseen spatial cells. Our results also show that GeST can extract spatial niche embeddings in a zero-shot way and can be further fine-tuned for spatial annotation tasks. Furthermore, GeST can simulate gene expression changes in response to perturbations of cells within spatial context, closely matching existing experimental results. Overall, GeST offers a powerful generative pre-training framework for learning spatial contexts in spatial transcriptomics.

bioinformatics↗

bioRxiv2025

SpatialAgent: An Autonomous AI Agent for Spatial Biology

Advances in AI are transforming scientific discovery, yet spatial biology, a field that deciphers the molecular organization within tissues, remains constrained by labor-intensive workflows. Here, we present SpatialAgent, an autonomous AI agent for spatial biology research. SpatialAgent couples large language models with a Plan-Act-Conclude architecture, dynamic tool and skill retrieval, multimodal interpretation, and verification modules that audit generated claims. It supports the full discovery loop, from gene-panel design and multimodal annotation to trajectory inference, cell-cell communication analysis, imputation, and hypothesis generation. Across human and mouse brain, heart, tonsil, colon, and prostate datasets, SpatialAgent outperformed established computational baselines and matched or surpassed expert scientists in key tasks. In open-ended case studies, it recovered known tissue organization and generated spatially grounded hypotheses. In a prospective mouse prostate cancer Xenium study, it designed a compact 100-gene add-on panel that profiled 4.2 million cells across 21 samples, improved cell-type and malignant-state prediction, and captured spatially structured tumor and microenvironment programs. SpatialAgent establishes a framework for autonomous and collaborative discovery in spatial biology.

bioinformatics↗

bioRxiv2025

Multi-modal Diffusion Model with Dual-Cross-Attention for Multi-Omics Data Generation and Translation

Single-cell multi-omics technologies offer unprecedented opportunities to decipher complex cellular mechanisms. To overcome experimental limitations in scale, cost, and coverage, powerful computational methods are essential for integrating diverse data modalities and generating high-fidelity in-silico data. In this paper, we present scDiffusion-X, a latent diffusion model specifically designed for this purpose. The core innovation is a Dual-Cross-Attention (DCA) module that adaptively learns intricate, hidden relationships between different molecular modalities, offering a more flexible and interpretable approach than traditional integration strategies. Extensive benchmarking experiments demonstrate that scDiffusion-X excels at generating realistic multi-omics data, preserving cellular heterogeneity and global data structures with excellent scalability. Distinct from existing multi-omics simulators, scDiffusion-X uniquely enables high-fidelity modality translation by predicting one molecular modality from another and provides robust uncertainty quantification. Beyond data generation, we designed a gradient-based interpretation framework to transform DCA module into a discovery tool, enabling inference of comprehensive cell-type-specific heterogeneous gene regulatory networks (GRNs). By integrating state-of-the-art generative modeling with deep biological interpretability, scDiffusion-X serves as a powerful tool for dissecting regulatory relationships, predicting perturbation responses, poised to accelerate discovery in single-cell multi-omics research.

bioinformatics↗

bioRxiv2024

PathOmCLIP: Connecting tumor histology with spatial gene expression via locally enhanced contrastive learning of Pathology and Single-cell foundation model

Tumor morphological features from histology images are a cornerstone of clinical pathology, diagnostic biomarkers, and basic cancer biology research. Spatial transcriptomics, which provides spatially resolved gene expression profiles overlaid on histology images, offers a unique opportunity to integrate morphological and expression features, thereby deepening our understanding of tumor biology. However, spatial transcriptomics experiments with patient samples in either clinical trials or clinical care are costly and challenging, whereas histology images are generated routinely and available for many legacy prospective cohorts of disease progression and outcomes in well-annotated cohorts. Inferring spatial transcriptomics profiles computationally from these histology images would significantly expand our understanding of tumor biology, but paired data for training multi-modal spatial-histology models remains limited. Here, we tackle this challenge by incorporating performant foundation models pre-trained on massive datasets of pathology images and single-cell RNA-Seq, respectively, which provide useful embeddings to underpin multi-modal models. To this end, we developed PathOmCLIP, a model trained with contrastive loss to create a joint-embedding space between a histopathology foundation model and a single-cell RNA-seq foundation model. We incorporate a set transformer to gather localized neighborhood tumor architecture following contrastive training, which further enhances performance and is necessary to obtain robust results. We validate PathOmCLIP across five tumor types and achieve significant performance improvements in gene expression prediction tasks over other methods. PathOmCLIP can be applied to many archived histology images, unlocking valuable clinical information and facilitating new biomarker discoveries.

bioinformatics↗