Biology subjects

Tang, M.

Publications and source records attributed to Tang, M..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

SpoT-mediated reduction of (p)ppGpp levels promotes Ralstonia pseudosolanacearum adaptation to both plant xylem and legume nodules

During evolution, bacteria have developed the ability to interact intimately with eukaryotic hosts. These interactions span a dynamic continuum ranging from pathogenicity to mutualism, along which bacteria can rapidly evolve and shift their lifestyles. However, the molecular mechanisms that enable bacteria to adapt to new hosts and to transition between distinct interaction modes remain poorly understood. Here, using a unique combination of two independent evolution experiments, we identified and characterized parallel adaptive mutations in spoT, which encodes the bifunctional (p)ppGpp synthetase-hydrolase. These mutations promote the adaptation of the plant pathogen Ralstonia pseudosolanacearum to two distinct plant-associated environments and two distinct lifestyles, the xylem of both susceptible and tolerant host plants as a pathogen and the root nodules of a legume as a symbiont, without compromising virulence on susceptible hosts. These mutations enhance the utilization of multiple carbon and nitrogen sources, including substrates known to be abundant in xylem sap, and increase bacterial exponential growth rate in minimal medium, suggesting reduced basal (p)ppGpp levels. Assessment of a strain deficient in SpoT synthetase activity confirmed that lowering basal (p)ppGpp levels is adaptive in both plant environments. Together, our findings reveal that fine-tuning intracellular (p)ppGpp concentrations represents an efficient strategy for optimizing bacterial adaptation to complex host-associated environments.

microbiology↗

bioRxiv2025

GPTAnno: Ontology-tree-guided hierarchical cell type annotation based on GPT models for single-cell data

Cell type annotation is critical for interpreting single-cell transcriptomic data but remains challenging due to uncertain cellular clustering granularity and inconsistent labeling across studies. Here we present GPTAnno, an automated, ontology-tree-guided, uncertainty-aware, hierarchical cell type annotation method based on GPT models. GPTAnno directly handles gene expression matrices, integrates multi-resolution clustering with large language model reasoning constrained by the cell ontology to produce standardized, ontology-aware, and reproducible annotations with automatic resolution selection. GPTAnno selects optimal clustering resolutions based on the annotation distance on the ontology tree and quantifies annotation uncertainty to flag ambiguous clusters for expert review. Benchmarking across twelve large-scale datasets demonstrates GPTAnnos superior accuracy on annotating cell types across various species, tissues, and disease contexts against existing methods. Implemented in R and Python with Seurat and Scanpy compatibility, GPTAnno allows simple inputs to streamline the reproducible annotation, considerably reducing human efforts in repeated reclustering, assigning and examining the labels.

genomics↗

bioRxiv2025

An Integrated Single-Cell Atlas Reveals Hepatic Stellate Cell Heterogeneity and Spatiotemporal Dynamics after Liver Injury

BackgroundHepatic stellate cells (HSCs) orchestrate fibrosis-free repair after acute liver injury (ALI) and sustain fibrogenesis in chronic liver injury (CLI). However, the heterogeneity and spatiotemporal dynamics of HSCs across these different injury models remain elusive. This study sought to construct a cross-etiological HSC atlas to delineate HSC states and transitions during liver injury. MethodsWe integrated 86,072 single-cell transcriptomes from 84 mouse samples across four etiologies and validated findings using multi-omics data from 277 mouse and 798 human samples. Cellular dynamics were characterized through clustering, trajectory inference, spatial analysis, and multicellular coordination network analysis. Experimental validation included liver injury models and gain-of-functional assays in primary mouse HSCs and LX-2 cells. ResultsWe established a cross-etiological, spatiotemporally resolved HSC atlas comprising 11 subpopulations. Trajectory analysis delineated a continuous quiescence-activation-attenuation (QAA) trajectory, recapitulating the in vivo full spectrum of state transitions and being supported by sequential pathway activation validated in vitro. In ALI, HSCs spatiotemporally completed the QAA trajectory around injury zones, whereas collagen-producing S100a6 HSCs pathologically accumulated in mouse and human CLI due to trajectory dysregulation. Notably, the atlas identified a previously unrecognized apoptosis-prone Mrc2 HSC subtype strongly co-localized with p53 in both mice and humans. Overexpression of transcription factors confirmed that Hbp1, Tbx20, Atoh8, and Plagl1 enriched in Mrc2+ HSCs promoted HSC apoptosis. Finally, we revealed the microenvironment of distinct cellular modules which coordinated HSC progression along the QAA trajectory. A 531-gene signature derived from the inflammatory-fibrotic cellular module significantly correlated with fibrosis stage and hepatocellular carcinoma risk in human cohorts. ConclusionsWe established a comprehensive HSC atlas and delineates HSC heterogeneity and spatiotemporal dynamic across etiologies. Dysregulation of the QAA trajectory underlies fibrotic progression, providing a resource for identifying antifibrotic targets.

cell biology↗

bioRxiv2025

Ancestral Sequences Cannot be Accurately Reconstructed via Interpolation in a Variational Autoencoder's Latent Space

Standard methods for ancestral sequence reconstruction (ASR) rely on substitution models for the residues in a biological sequence and assume independent evolution across these sites, ignoring the epistatic interactions that shape molecular evolution. In contrast, deep learning models like variational autoencoders (VAEs) can learn low-dimensional representations ("embeddings") of sequences in a protein family that may implicitly handle these dependencies, raising the possibility of performing more accurate ASR by interpolating between extant sequence embeddings within the VAE's latent space. In this study, we test this hypothesis by developing and evaluating a VAE-based ASR pipeline. Benchmarking this approach against established likelihood-based and parsimony methods using various simulations of protein evolution, including scenarios with and without epistasis, we find that the VAE-based approach is consistently and significantly outperformed by standard methods, even in epistatic regimes where it was hypothesized to have an advantage. We further show that this failure is not due to a lack of phylogenetic structure in the latent space, which does contain evolutionary signal. Rather, the primary limitation is the information loss inherent to the autoencoding process: the VAE's decoder cannot generate sequences with sufficient fidelity for the precise demands of ASR.

evolutionary biology↗

bioRxiv2025

Pervasive context-dependent effects in the genetic architecture of complex and quantitative traits revealed by a powerful multiparent mapping population in yeast

The genetic dissection of complex traits remains a major challenge in basic and biomedical research, but is essential for understanding the molecular pathways that shape phenotypic variation and for developing predictive models of trait and disease susceptibility. Here, we leverage a novel multiparent mapping population of budding yeast, CYClones, comprising 9,344 haploid strains derived from eight genetically diverse founders, to identify quantitative trait loci (QTL) and systematically investigate the genetic architecture of growth rates across ten environmental conditions. In total, we identified 349 QTL (ranging from 19 to 49 QTL per growth condition) that explained between 61% and 100% of narrow sense heritability across traits. The high power and resolution of CYClones revealed that growth traits exhibited distinct, condition-specific genetic architectures with extensive allelic heterogeneity, where a QTL was the result of multiple tightly linked causal variants. We also observed pleiotropy among QTL with complex, trait-dependent allele effects that are also consistent with allelic heterogeneity. Genetic complexity varied widely, with some traits showing nearly Mendelian architectures, while others were highly polygenic. Introgressed loci played a prominent role in the landscape of growth rate QTL, including one that is localized to a 2.4 kb interval in the PCA1 cadmium transporter and explains 72% of variation in cadmium resistance and another large-effect QTL that alters growth on galactose via complex non-additive interactions with the transcriptional regulator GAL3. In both cadmium and galactose conditions, we show that allelic variation at a small number of loci stratifies the population into regulatory or physiological subgroups, each with distinct genetic architectures, a phenomenon we term allele-dependent stratification. Collectively, our results provide novel insights into the genetics of growth rates in budding yeast, the architectural features of genetic complexity, and demonstrate that CYClones is a powerful platform for revealing the molecular basis of complex trait variation.

genetics↗

bioRxiv2025

scRetinaDB: A Comprehensive Database of Single-Cell and Spatial Omics from Cross-Species Retinas

The retina is essential for encoding visual signals, and its dysregulation can lead to retinal diseases. Recent advances in single-cell and spatial sequencing technologies have yielded extensive omics data from retinal tissues across species and biological conditions. However, existing retinal omics data are dispersed across various repositories without uniform processing, which limits integrative analysis. To address this we developed scRetinaDB (https://casapp.dnayun.com/scretina/), a comprehensive resource that aggregates single-cell RNA sequencing (scRNA-seq), single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq), and spatial RNA sequencing (spRNA-seq) data from retinas across species and diverse biological conditions. The database comprises over 2.79 million retinal cells collected from 453 scRNA-seq datasets spanning 34 studies, 17 species and 27 biological conditions. For each species, these scRNA-seq datasets were integrated to construct a retinal cell atlas. In addition, scRetinaDB also contains 107 scATAC-seq and 18 spRNA-seq datasets from human and mouse retinas. The scRetinaDB website provides four major modules separately for browsing species-specific omics data, searching cross-species omics profiles, performing analyses of cell type annotation and cell similarity analysis, and downloading preprocessed multi-omics datasets. Overall, scRetinaDB is a valuable resource for retinal single-cell and spatial omics, advancing cross-species studies particularly in vision research.

bioinformatics↗

bioRxiv2025

Phosphodiesterase type 1 regulates adenosine 2A receptor-associated cAMP signaling at the plasma membrane to increase myocardial contractility

BACKGROUNDPhosphodiesterase type 1 (PDE1) inhibition exerts inodilatory effects in pre-clinical models and human heart failure patients with reduced ejection fraction (HFrEF). PDE1 hydrolyzes cyclic nucleotide cAMP in a soluble but not microsomal fraction of the human myocardium. PDE1 may exert domain-specific effects, but the mechanism whereby PDE1 compartmentalization induces an inotropic change remains unknown. We sought to elucidate PDE1 regulation of cAMP and contractility. METHODSPharmacologic modulators of PDEs or G-protein coupled receptors (GPCRs) were used to study PDE1 signaling mechanisms in healthy and failing guinea pig hearts. Tissue fractionation examined the localization of different PDE types. Live cell imaging experiments assessed cytosolic and sarcolemmal membrane cAMP level ([cAMP]) and protein kinase A (PKA) activity changes in cells transduced with Forster resonance energy transfer (FRET) biosensors. Sarcomere length and intracellular calcium changes monitored contractile changes in electrically paced cells. Coronary flow and left ventricular (LV) developed pressure were measured in ex vivo Langendorff perfusion heart studies. RESULTSPDE1 isoforms were found not only in the soluble, but also in the microsomal fractions of the guinea pig heart. PDE1 hydrolysis of cAMP was greater at the sarcolemma compared to the cytosol. PDE1 specifically regulated a pool of cAMP associated with the Gs protein coupled receptor adenosine 2A receptor (A2AR) at the sarcolemma, without activating PKA. A2AR/PDE1 regulation induced positive inotropic and lusitropic changes in healthy and failing guinea pig cardiomyocytes and in the myocardium ex vivo. CONCLUSIONSPDE1 is the major regulator of cAMP pools generated by A2AR activation at the sarcolemma. Functionally, this regulation induces inotropic and lusitropic effects in cardiomyocytes and at the whole organ level. Thus, PDE1 is compartmentalized at the membrane with A2AR, and this regulation determines cardiomyocyte contractility in healthy and failing hearts.

physiology↗

bioRxiv2025

FUSED: CROSS-DOMAIN INTEGRATION OF FOUNDATION MODELS FOR CANCER DRUG RESPONSE PREDICTION

AO_SCPLOWBSTRACTC_SCPLOWAI-driven methods for predicting drug responses hold promise for advancing personalized cancer therapy, but cancer heterogeneity and the high cost of data generation pose substantial challenges. Here we explore the transfer learning capability and introduce FUSED (Fusion of Foundation Model Embeddings for Drug Response Prediction), a novel architecture for cross-domain foundation model (FM) integration. By systematically benchmark FMs across two domains - molecular FM for drugs and single-cell FM for cell lines, we demonstrate that integrating single-cell FMs substantially reduces the number of input features required for cell line representation. Among FMs, Molformer significantly outperforms ChemBERTa, and scGPT surpasses scFoundation in predictive accuracy and training stability. Moreover, integrating single-cell FMs improves performance in both drug-known and leave-one-drug-out scenarios. These findings highlight the potential of cross-domain FM integration for more efficient and robust drug response prediction.

bioinformatics↗

bioRxiv2025

Functional Profiling of 2,193 ASS1 Missense Variants: Insights into Variant Pathogenicity and Epistatic Interactions in Citrullinemia Type I

Sequence variants in the urea cycle gene argininosuccinate synthase (ASS1) cause Citrullinemia type 1 (CTLN1), a rare autosomal recessive disease. Mechanistically, reduction in argininosuccinate synthetase (ASS) enzyme activity impairs the urea cycle, leading to an accumulation of citrulline and neurotoxic ammonia. Disease severity varies according to the degree of enzyme impairment, ranging from severe neonatal forms (classic citrullinemia) to milder, late-onset forms that may manifest in childhood or adulthood. We established a high-throughput yeast functional assay of human ASS and individually measured the impact of 2193 amino acid substitutions, representing 90% of all single nucleotide variant (SNV)-accessible substitutions. When benchmarked against existing clinical variant annotation, our assay distinguishes known benign variants from strong loss of function pathogenic variants, enabling identification of a functional score threshold below which variants show clinically relevant impairment of ASS activity. Using the ACMG OddsPath framework, our assay meets PS3_supporting criteria for pathogenicity classification and achieves full PS3-level strength when variants observed as homozygotes in other primates are used as benign proxies for calibration. These results provide direct functional evidence to inform reclassification of ASS1 missense variants, supporting their clinical interpretation and diagnostic utility. Mapping functional scores onto the protein structure, we confirmed that residues involved in catalysis are highly sensitive to substitution. In addition, we identified residues from adjacent subunits of the ASS homotetramer that form compound active sites. Assaying these positions revealed a capacity for intragenic complementation consistent with a variant sequestration model: a form of positive epistasis in which deleterious variants from different subunits are sequestered into only a subset of active sites, restoring function in the remaining variant-free sites. The discovery of intragenic complementation in ASS reveals a novel mode of functional interaction with clinical implications for interpreting variant combinations in heterozygous individuals.

genetics↗

bioRxiv2025

Gliomas phenocopy an inborn error of metabolism to drive neuronal activity and tumor growth

The metabolic hallmarks of high-grade glioma (HGG) are not fully understood. Human brain tissue metabolomics revealed that the creatine synthesis pathway intermediate guanidinoacetate (GAA) accumulated [~]100-fold in HGGs relative to controls, which was caused by imbalanced activities of enzymes in this pathway. Glioma cells secreted GAA rather than using it to produce creatine, implicating an unexpected function. GAA accumulates in GAMT deficiency, an inborn error of metabolism, and elevates neuronal excitability. Neuronal excitability is also increased in glioma and drives tumor growth through neuron-glioma interactions. We hypothesized that glioma-generated GAA excites surrounding neurons. Indeed, GAA induced neuronal hyperactivity by activating GABAA receptors and causing depolarizing GABA currents in glioma-associated neurons with dysregulated chloride homeostasis. Depleting tumoral GAA decreased electrochemical activity, neuron-glioma interactions, and tumor aggressiveness. Our findings unveil a new mechanism linking cancer metabolism with cancer neuroscience and leverage human genetics to nominate GAA synthesis as a target in gliomas.

cancer biology↗

bioRxiv2025

The Annotated Blueprint: Integrated Functional Genomic Resources for a Model Tetraploid Wheat Triticum turgidum cv. Kronos

Triticum turgidum cv. Kronos is a tetraploid wheat cultivar that underpins one of the richest community platforms for functional genomics. Over the past decade, about 3,000 exome- and promoter-capture datasets, linked to mutagenized seed stocks, and transcriptomic and phenotypic resources have accumulated, yet the absence of a reference genome has constrained their impact. Here, we present a chromosome-scale reference genome of Kronos with high-confidence annotations, including manual curation of over 1,000 disease resistance (NLR) genes. This reference revealed previously hidden NLR diversity and clarified their genomic organization at chromosomal ends. Re-analysis of exome- and promoter-capture datasets enabled high-resolution mutation discovery in genes and regulatory regions that were previously inaccessible, uncovering the full standing variation present in Kronos mutant lines. We further re-curated transcriptomic and small RNA datasets, generating improved, genome-wide maps of microRNAs and phasiRNAs important for wheat development. Collectively, these resources elevate Kronos to reference quality and establish it as a versatile platform for functional and translational wheat research.

plant biology↗

bioRxiv2025

Diagnosing linearity along the carbon cascade in terrestrial biosphere models

Elevated carbon dioxide (eCO2) fertilises photosynthesis, driving an increase in terrestrial gross primary production (GPP). However, it is unclear how effectively increased GPP propagates along the "carbon (C) cascade" to increase net primary production (NPP) and vegetation C stocks (Cveg) in different plant compartments. Vegetation models simulate divergent C cycle projections and have been criticised for being overly photosynthesis-driven (source-driven), neglecting processes that lead to non-linear behaviour in response to the GPP increase, which may attenuate (or amplify) changes in NPP and vegetation C stocks. Here, we introduce an analytical framework to diagnose linearity (L) of the land C cycle as the ratio of relative changes in linked fluxes and pools and apply it to outputs from 16 models of the TRENDY v11 ensemble. We found widely varying patterns in L across models and for the different links. Six models showed a clear dominance of larger relative changes in NPP than in GPP in global simulations (LNPP:GPP >1 for >60% of gridcells), indicating increased carbon use efficiency under eCO2. Only three models had LNPP:GPP < 1 for >60% of gridcells. Four models showed a clear dominance of larger relative changes in steady-state Cveg than in NPP, while five models showed an opposite pattern - in both cases with a large spread of LCveg*:NPP across gridcells within models. Three models showed a larger relative increase in root C than in Cveg, while two models showed a clear dominance of the opposite pattern. Widely differing distributions of L across models and links reveal a strong influence of alternative process representations (nonlinear behaviour) in individual models. However, for all links, L deviations from 1 were roughly balanced across the model ensemble, leading to an overall linear behaviour of terrestrial C cycle representations.

ecology↗

bioRxiv2025

An AI-based and coding-free protocol for forests Leaf Area Index (LAI) calculation

O_LISeasonal and spatial variations in leaf area index (LAI) are challenging to detect in tropical forests due to dynamic lighting conditions and the subtle differences in the variation. Many existing LAI software tools offer one-click processing of all images through auto-threshold segmentation (e.g., HemispheR, HemiPy and Hemisfer), but they produce results with large discrepancies. Some software (e.g. CAN-EYE) requires manual tuning of each image, making large-scale analysis impractical. C_LIO_LIWe analysed 19,000 images from four tropical forest subtypes and found that using coding-free AI software to process hemispherical images can significantly improve the consistency of leaf-sky segmentation, thereby enhancing LAI outcomes. C_LIO_LIThe results show that replacing the auto-threshold with AI substantially reduced inter-software disagreement and delineated correct seasonal and spatial patterns. CAN-EYE was able to identify seasonal patterns but produced less accurate results than the CAN-EYE-AI integrated approach due to subjective user bias. C_LIO_LIThe high consistency achieved through AI integration enables reliable cross-site and cross-operator comparisons. As users can customise the AI model according to local images and combine the AI model with other LAI software, our integrated, affordable, and coding-free method offers wide applicability and high consistency of LAI measurements, facilitating the advancement of tropical forest monitoring and research. C_LI Data/Code for peer review statementOne of the key features of this method is coding-free. The method is explained in Protocolv20251118.docx. We have uploaded R codes for drawing figures in a zip pack. These codes and the protocol will be deposited in the Zenodo (or figshare) database under accession link [TBC]. Since Zenodo allows authors to archive updated versions after publication, we may update the protocol by uploading a revised version to Zenodo. Please check the Zenodo archive for any new versions. In the protocol, we note that users can use Image_conversion_20220407.m and lets_change_values.R instead of the Renormalise function of ilastik to modify values in the classification output images. These codes are not essential for users following our protocol, but could be useful for integrating ilastik with other LAI software not covered in this paper. Additionally, the protocol mentions that Gather_LAI_fapar_from_caneye.R can be used to consolidate output Excel files, eliminating the need to manually open each file. Field measurements of LAI and GCC are available on request.

ecology↗

bioRxiv2025

Circulating tumor cell heterogeneity and MMP9 expression in clinical correlation with breast cancer

ObjectivesThis study aimed to investigate the association between the heterogeneity of circulating tumor cell (CTC) subtypes and the expression patterns of matrix metalloproteinase 9 (MMP9) with clinicopathological parameters in breast cancer patients. The distribution of CTC subtypes, MMP9 expression levels, and their influence on tumor progression. MethodsA total of 303 breast cancer patients were enrolled in this study. The association between CTCs subtypes in peripheral blood, MMP9 expression and clinical or pathological data of patients were analyzed. ResultsThe number of mesenchymal CTCs increased significantly with the number of lymph node metastases (P < 0.05), and high MMP9 expression was positively correlated with lymph node involvement (P < 0.05). The proportion of intermediate to high MMP9 expression was significantly higher in mixed CTCs compared to epithelial and mesenchymal CTCs (P < 0.05). High MMP9 expression in tumor tissues was associated with increased CTCs production and induction of epithelial-mesenchymal transition (EMT) (P < 0.05), as well as an elevated number of MMP9-positive CTCs (P < 0.05). ConclusionsThese findings suggest that MMP9 is positively associated with CTCs production and EMT process, and MMP9 may contribute to tumor metastasis by promoting CTC shedding and EMT by degrading the extracellular matrix. We recommend compliance with the Mesh guidelines: https://www.ncbi.nlm.nih.gov/mesh/

cancer biology↗

bioRxiv2025

Deepening imaging-based spatial proteomics at high spatial resolution through controlled tissue resizing and in-situ bottom-up mass spectrometry

Understanding the spatial organization of proteins in tissues is essential for elucidating biological function, but current proteomic imaging methods face limitations in resolution, sensitivity, and multiplexing capability. Here, we present IMPACT (Imaging Mass Spectrometry for Proteome Analysis on Compressed Tissue), a hydrogel-based workflow that integrates controlled vertical tissue compression following expansion to enhance protein detection while improving spatial resolution in mass spectrometry imaging (MSI). Through a 6-fold lateral expansion and 100-fold thickness reduction, analytes were concentrated at the surface to improve detection and enable analysis of thick specimen. Coupled with optimized in-gel digestion using hybrid enzymes, IMPACT increased protein identifications 3-fold compared to conventional MSI while maintaining spatial fidelity. We demonstrate IMPACTs utility in mapping low-abundance proteins and resolving fine anatomical structures. Orthogonal validation with immunofluorescence and integration with spatial transcriptomics further reveal protein-RNA distribution discrepancies, highlighting the need for direct proteomic imaging. IMPACT bridges critical gaps in spatial proteomics by enabling high-resolution, high-sensitivity, and multiplexed untargeted protein imaging, opening new avenues for tissue biology and biomarker discovery.

biophysics↗

bioRxiv2025

Mouse Bio-behavioral Phenotyping Using a Digital Homecage Framework for Long-timescale, High-resolution, and Multi-factor Data Collection and Analytics

MOTIVATION Long-term monitoring of behavioral and physiological processes is critical for understanding complex brain-based phenomena and disorders that develop over extended periods, such as chronic stress, circadian disorders, and metabolic conditions. While digital phenotyping is available in humans using smart devices, there remains a deficit in rodent models. To address this lack of long-term rodent phenotyping, we introduce the "Digital Homecage" (DHC) system. Our system uses accessible components and is designed for seamless integration with brain recording technologies. The Digital Homecage (DHC) allows uninterrupted, long-timescale recording of more than 20 behavioral metrics in single-housed mice, captured at sub-second resolution via video, operant interactions, and wheel-running data. This report demonstrates the DHCs capacity to enable continuous, automated tracking of behaviors like actigraphy, sleep, grooming, and food choice options over weeks, thereby opening up new avenues for longitudinal analyses of chronic conditions. Data collected reveal circadian patterns in multiple spontaneous behaviors, aligning with known nocturnal tendencies. The systems potential to facilitate groundbreaking observations on the long-term behavioral correlates of various neuropsychiatric syndromes is aided by open-source software and relatively low cost. The DHC sets the stage for community-driven innovation, potentially transforming our approach to studying complex traits of brain function and behavior in laboratory settings.

animal behavior and cognition↗

bioRxiv2025

RNAchat: Integrating machine learning algorithms to identify metapathways based on clinical and multi-omics data

BackgroundUnderstanding the global interactions among cellular pathways and types is essential for unraveling the complex biological mechanisms that underlie various diseases. While many existing studies have primarily focused on omics data, they often fail to integrate clinical heterogeneity, which is crucial for a comprehensive understanding of disease biology. Although significant progress has been made in researching crosstalks between inter-pathways and inter-cells, and tools like CellChat have emerged to analyze these interactions, they have notable limitations. These include the assumption that mRNA expression directly reflects protein activity, a heavy reliance on pre-established ligand-receptor interactions, and an inability to connect "chats" with clinical phenotypes. Furthermore, existing methods have limitations in distinguishing between paracrine signaling (interactions between different cell types) and autocrine signaling (self-signaling within the same cell type). These limitations hinder the ability to translate molecular insights into phenotypic outcomes, thereby limiting clinical applicability. To provide another solution for these challenges, we introduce RNAchat, a novel approach that utilizes machine learning techniques to identify metapathways for crosstalks among pathways at not only the bulk level but also cell types at the single-cell level, incorporating both clinical and multi-omics data. ResultsRNAchat provides another solution by providing an interactive, reproducible platform designed to analyse metapathways between inter-pathways and inter-cell types using multi-omics level data in a clinical context. This tool enables researchers to integrate diverse datasets, conduct exploratory analyses, and identify cooperated pathways. The platform facilitates hypothesis generation and produces intuitive visualizations to support further investigation. As a proof of concept, we applied RNAchat to connect omics data to pain, drug resistance in rheumatoid arthritis (RA), disease severity to RA and Heart Failure (HF). We discovered cooperated molecular pathways associated with different treatments using SHAP (Shapley Additive Explanations) values. Then, we demonstrated how to use it to find cellular interactions at single cell level. Last but not the least, we further extended crosstalk exploration between host and parasites and found potential novel important pathway interaction not mentioned by previous studies. ConclusionWe introduce RNAchat, a computational platform designed to identify pathway communications with clinical and multi-omics data in real-time. This tool supports both user-generated and publicly available datasets, offering a robust solution and enhancing our understanding of complex diseases such as RA and HF. The platform is available at https://github.com/tangmingcan/RNAchat.

bioinformatics↗

bioRxiv2025

RNAcompare: Integrating machine learning algorithms to unveil the similarities of phenotypes based on clinical, multi-omics using Rheumatoid Arthritis and Heart Failure as Case Studies

BackgroundGene expression analysis is crucial for understanding the biological mechanisms underlying patient subgroup differences. However, most existing studies focus primarily on transcriptomic data while neglecting the integration of clinical heterogeneity. Although batch correction methods are commonly used, challenges remain when integrating data across different tissues, omics layers, and diseases. This limitation hampers the ability to connect molecular insights to phenotypic outcomes, thereby restricting clinical translation. Furthermore, the technical complexity of analysing large, heterogeneous datasets poses a barrier for clinicians. To address these challenges, we present RNAcompare, a development of RNAcare, employing machine learning techniques to integrate clinical and multi-omics data seamlessly. ResultsRNAcompare overcomes these challenges by providing an interactive, reproducible platform designed to analyse multi-omics data in a clinical context. This tool enables researchers to integrate diverse datasets, conduct exploratory analyses, and identify shared patterns across patient subgroups. The platform facilitates hypothesis generation and produces intuitive visualizations to support further investigation. As a proof of concept, we applied RNAcompare to connect omics data to pain, fatigue, drug resistance in rheumatoid arthritis (RA) and disease severity to RA and Heart Failure (HF). Our analysis reduced selection bias and managed heterogeneity by identifying key contributors to treatment variability. We discovered shared molecular pathways associated with different treatments. Using SHAP (Shapley Additive Explanations) values, we successfully classified patients into three subgroups based on age, and subsequent analyses confirmed these age-related patterns. Additionally, we uncovered hidden patterns influencing pain and disease severity across different tissues, omics layers, and diseases. Notably, by integrating Causal Forests and Double Machine learning with clinical phenotypes, RNAcompare provides a novel approach to bypass traditional batch correction methods. ConclusionWe introduce RNAcompare, a computational platform designed to compare clinical and multi-omics data across diverse patient cohorts in real-time. This tool supports both user-generated and publicly available datasets, offering a robust solution for identifying phenotypic similarities and enhancing our understanding of complex diseases such as RA and HF. The platform is available at https://github.com/tangmingcan/RNAcompare.

bioinformatics↗