Biology subjects

Madgwick, M.

Publications and source records attributed to Madgwick, M..

12 recordsLinked to original sources

bioRxiv2026

Macrophage signature-based prediction of cancer treatment response using MIL-attention

Predicting immunotherapy response from single-cell data remains difficult due to patient-level labels, extreme class imbalance, and highly heterogeneous macrophage states. We present a Multiple Instance Learning (MIL) framework that treats each patient as a bag of macrophage embeddings derived from a single-cell RNA foundation model. The architecture incorporates an attention-based pooling mechanism with reduced model complexity, dropout-enhanced regularization and explicit attention penalties to improve stability in small-sample regimes. To address imbalanced clinical datasets, MIL outputs are optimized with a combined focal loss and supervised contrastive objective that simultaneously sharpens class boundaries and improves representation clustering. Across three cancer datasets, this approach outperforms pseudobulk aggregation, embedding baselines and standard MIL variants. Attention-weighted attribution and transcriptional regulatory analysis reveal distinct macrophage programs, interferon and antigen-presentation networks in responders versus hypoxia-linked regulatory modules in non-responders. This shows the potential of MIL to uncover predictive and mechanistically interpretable immune states.

bioinformatics↗

bioRxiv2026

A microbiota-responsive polyfunctional cytotoxic CD4⁺ T-cell state promotes mucosal inflammation in ulcerative colitis

Ulcerative colitis (UC) is characterised by chronic colonic inflammation with marked heterogeneity in disease severity and therapeutic outcomes. Here, we define a spatially organised, polyfunctional cytotoxic CD4 T-cell state associated with mucosal inflammation and adverse therapeutic outcomes in UC. Integrating ex vivo T-cell receptor stimulation with multi-cohort bulk and single-cell transcriptomics and multiparameter flow cytometry, we show that GZMB CD4 T cells are preferentially enriched in inflamed UC mucosa, but not peripheral blood, and co-express cytotoxic molecules, Th1- and Th17-associated cytokines and chemokines, and immunoregulatory receptors. Single-cell analyses implicate inflammatory cytokine and antigen-presentation signals in the acquisition or maintenance of this state. High-resolution spatial profiling localised this programme predominantly to Th17 cells, which were preferentially enriched within multicellular inflammatory and tertiary lymphoid structure-associated niches. Across independent patient cohorts, a transcriptional signature derived from this state increased with endoscopic disease severity and was associated with reduced response to anti-TNF and anti-IL-12/23p40 therapies. Adoptive transfer of Gzma/Gzmb-deficient rather than wild-type CD4 T cells into Rag2-deficient recipient mice markedly attenuated experimental colitis and abrogated the polyfunctional cytokine phenotype, demonstrating that granzyme-dependent effector activity is a key mechanism driving CD4+ T-cell-mediated intestinal inflammation. Finally, human host-microbiome analysis linked this programme to intestinal dysbiosis, while transfer of dysbiotic microbiota promoted the emergence of a corresponding state in vivo. Collectively, these findings define a microbiota-responsive, spatially organised polyfunctional cytotoxic CD4 T-cell programme that contributes to intestinal inflammation and is associated with disease severity and treatment resistance in UC.

immunology↗

bioRxiv2026

RNA foundation models enable generalizable endometriosis disease classification and stable gene-level interpretation

Endometriosis is a chronic inflammatory condition with significant diagnostic delays impacting one in ten reproductive age women worldwide. While machine learning (ML) models trained on transcriptomic data show promise for disease prediction, limited generalizability across independent patient cohorts has hindered clinical translation. Foundations models (FMs) pretrained on large-scale transcriptomic data offer promise to learn transferrable, biologically meaningful representations that could support cross-cohort predictions. We assembled a 12-cohort bulk RNA-seq benchmark (334 samples) and developed a computationally efficient pipeline to test whether FMs improve endometriosis classification, an approach not previously applied to this disease. Using AutoXAI4Omics with cohort-aware validation, we compared embeddings derived from five state-of-the-art RNA FMs against TPM baselines. In cross-cohort prediction, FM embeddings significantly improved performance, achieving a weighted F1-score of 0.83 vs. 0.68 for the baseline. To allow gene-level interpretation of FM embedding models, we introduce classified-aligned integrated gradients (CA-IG), an interpretability approach aligning gene-level attributions to the downstream classifier without end-to-end finetuning. CA-IG revealed a conserved set of predictive genes from FM embeddings across cohort-validation regimes, contrasting with unstable baseline explainability, suggesting that FM embeddings prioritized transferable disease-related signal over cohort-specific effects. These genes include novel candidates that converge on biologically plausible pathways for endometriosis.

bioinformatics↗

bioRxiv2024

AutoXAI4Omics: an Automated Explainable AI tool for Omics and tabular data

Machine learning (ML) methods have the potential of detailed insights of complex biological systems and today are increasingly used to analyse omics data for tasks such as the discovery of novel biomarkers and phenotype prediction. It can be extremely beneficial and powerful for scientists, domain experts, to easily run sophisticated, robust, and interpretable ML pipelines without the need for an in depth understanding of the code needed to train, tune, optimise ML algorithms. They can then focus on the biological interpretation and validation of the results and insights generated by ML models. Here, we present an entirely automated open-source explainable AI tool, AutoXAI4Omics, that performs classification and regression tasks from omics and tabular numerical data. AutoXAI4Omics accelerates scientific discovery by automating processes and decisions made by AI experts, e.g., selection of the best feature set, hyper-tuning of different ML algorithms and selection of the best ML model for a specific task and dataset. Prior to ML analysis AutoXAI4Omics incorporates feature filtering options that are tailored to specific omic data types. Moreover, the insights into the predictions that are provided by the tool through explainability analysis highlight associations between omic feature values and the targets under investigation e.g., predicted phenotypes, facilitating the discovery of actionable insights. AutoXAI4Omics is at: https://github.com/IBM/AutoXAI4Omics. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=188 HEIGHT=200 SRC="FIGDIR/small/586460v1_ufig1.gif" ALT="Figure 1"> View larger version (36K): org.highwire.dtl.DTLVardef@366327org.highwire.dtl.DTLVardef@a7b559org.highwire.dtl.DTLVardef@7319aforg.highwire.dtl.DTLVardef@9b5030_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2023

AutophagyNet: High-resolution data source for the analysis of autophagy and its regulation

Autophagy is a highly-conserved catabolic process eliminating dysfunctional cellular components and invading pathogens. Autophagy malfunction contributes to disorders such as cancer, neurodegenerative and inflammatory diseases. Understanding autophagy regulation in health and disease has been the focus of the last decades. We previously provided an integrated database for autophagy research, the Autophagy Regulatory Network (ARN). For the last seven years, this resource has been used by thousands of users. Here, we present a new and upgraded resource, AutophagyNet. It builds on the previous database but contains major improvements to address user feedback and novel needs due to the advancement in omics data availability. AutophagyNet contains updated interaction curation and integration of over 280,000 experimentally verified interactions between core autophagy proteins and their protein, transcriptional and post-transcriptional regulators as well as their potential upstream pathway connections. AutophagyNet provides annotations for each core protein about their role: 1) in different types of autophagy (mitophagy, xenophagy, etc.); 2) in distinct stages of autophagy (initiation, elongation, termination, etc); 3) with subcellular and tissue-specific localization. These annotations can be used to filter the dataset, providing customizable download options tailored to the users needs. The resource is available in various file formats (e.g., CSV, BioPAX and PSI-MI), and data can be analyzed and visualized directly in Cytoscape. The multi-layered regulation of autophagy can be analyzed by combining AutophagyNet with tissue- or cell type-specific using (multi-)omics datasets (e.g. transcriptomic or proteomic data). The resource is publicly accessible at http://autophagynet.org.

bioinformatics↗

bioRxiv2021

Multi-layered networks of SalmoNet2 enable strain comparisons of the Salmonella genus on a molecular level

Serovars of the genus Salmonella primarily evolved as gastrointestinal pathogens in a wide range of hosts. Some serotypes later evolved further, adopting a more invasive lifestyle in a narrower host range associated with systemic infections. A system-level knowledge of these pathogens has the potential to identify the complex adaptations associated with the evolution of serovars with distinct pathogenicity, host range and risk to human health. This promises to aid the design of interventions and serve as a knowledge base in the Salmonella research community. Here we present SalmoNet2, a major update to SalmoNet, the first multi-layered interaction resource for Salmonella strains, containing protein-protein, transcriptional regulatory and enzyme enzyme interactions. The new version extends the number of Salmonella genomes from 11 to 20, including strains such as S. Typhimurium D23580, an epidemic multidrug-resistant strain leading to invasive non-typhoidal Salmonella Disease (iNTS), and a strain from Salmonella bongori, another species in the Salmonella genus. The database now uses strain specific metabolic models instead of a generalised model to highlight differences between strains. This has increased the coverage of high-quality protein-protein interactions, and enhances interoperability with other computational resources by adopting standardised formats. The resource website has been updated with tutorials to help researchers analyse their Salmonella data using molecular interaction networks from SalmoNet2. SalmoNet2 is accessible at http://salmonet.org/. ImportanceMulti-layered network databases collate information from multiple sources, and are powerful both as a knowledge base and platform for analysis. Here we present SalmoNet2, an integrated network resource of 20 Salmonella strains, containing protein-protein, transcriptional regulatory, and metabolic interactions. Key improvements to the update include expanding the number of strains, strain-specific metabolic networks, an increase in high quality protein-protein interactions, community standard computational formats to help interoperability, and online tutorials to help users analyse their data using SalmoNet2.

systems biology↗

bioRxiv2021

CytokineLink: a cytokine communication map to analyse immune responses in inflammatory and infectious diseases

Intercellular communication mediated by cytokines is critical to the development of immune responses, particularly in the context of infectious and inflammatory diseases. By releasing these small molecular weight peptides, the source cells can influence numerous intracellular processes in the target cells, including the secretion of other cytokines downstream. However, there are no readily available bioinformatic resources that can model cytokine - cytokine interactions. In this effort, we built a communication map between major tissues and blood cells that reveals how cytokine-mediated intercellular networks form during homeostatic conditions. We collated the most prevalent cytokines from literature, and assigned the proteins and their corresponding receptors to source tissue and blood cell types based on enriched consensus RNA-Seq data from the Human Protein Atlas database. To assign more confidence to the interactions, we integrated literature information on cell - cytokine interactions from two systems immunology databases, immuneXpresso and ImmunoGlobe. From the collated information, we defined two metanetworks: a cell-cell communication network connected by cytokines; and a cytokine-cytokine interaction network depicting the potential ways in which cytokines can affect the activity of each other. Using expression data from disease states, we then applied this resource to reveal perturbations in cytokine-mediated intercellular signalling in inflammatory and infectious diseases (ulcerative colitis and COVID-19, respectively). For ulcerative colitis, with CytokineLink we demonstrated a significant rewiring of cytokine-mediated intercellular communication between non-inflamed and inflamed colonic tissues. For COVID-19, we were able to identify inactive cell types and cytokine interactions that may be important following SARS-CoV-2 infection when comparing the cytokine response with other viruses capable of initiating a cytokine storm. Such findings have potential to inform the development of novel, cytokine-targeted therapeutic strategies. CytokineLink is freely available for the scientific community through the NDEx platform and the project github repository (https://github.com/korcsmarosgroup/CytokineLink).

bioinformatics↗

bioRxiv2021

Reprogramming of the intestinal epithelial-immune cell interactome during SARS-CoV-2 infection

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) represents an unprecedented worldwide health problem. Although the primary site of infection is the lung, growing evidence points towards a crucial role of the intestinal epithelium. Yet, the exact effects of viral infection and the role of intestinal epithelial-immune cell interactions in mediating the inflammatory response are not known. In this work, we apply network biology approaches to single-cell RNA-seq data from SARS-CoV-2 infected human ileal and colonic organoids to investigate how altered intracellular pathways upon infection in intestinal enterocytes leads to modified epithelial-immune crosstalk. We point out specific epithelial-immune interactions which could help SARS-CoV-2 evade the immune response. By integrating our data with existing experimental data, we provide a set of epithelial ligands likely to drive the inflammatory response upon infection. Our integrated analysis of intra- and inter-cellular molecular networks contribute to finding potential drug targets, and suggest using existing anti-inflammatory therapies in the gut as promising drug repurposing strategies against COVID-19.

bioinformatics↗

bioRxiv2021

Extracellular vesicles produced by the human commensal gut bacterium Bacteroides thetaiotaomicron affect host immune pathways in a cell-type specific manner that are altered in inflammatory bowel disease

The gastrointestinal (GI) tract is inhabited by a complex microbial community, which contributes to its homeostasis. Disrupted microbiome can cause GI-related diseases, including inflammatory bowel disease, therefore identifying host-microbe interactions is crucial for better understanding gut health. Bacterial extracellular vesicles (BEVs), released into the gut lumen, can cross the mucus layer and access underlying immune cells. To study cross-kingdom communication between BEVs and host, we focused on the influence of BEVs, generated by Bacteroides thetaiotaomicron (VPI-5482), on host immune cells. Using single-cell RNA sequencing data and host-microbe protein-protein interaction networks, we examined the potential effect of BEVs on dendritic cells, macrophages and monocytes with particular focus on the Toll-like receptor (TLR) pathway. We identified biological processes affected in each immune cell type, and also cell-type specific processes (e.g myeloid cell differentiation). The TLR pathway analysis highlighted that BEV targets differ among cells and even between the same cells in healthy versus disease (ulcerative colitis) conditions. Our in silico findings were validated in BEV-monocyte co-cultures demonstrating the requirement for TLR4 in BEV-elicited NF-{kappa} B activation. This study demonstrates that both cell-type and health condition influence BEV-host communication. The results and the pipeline can facilitate BEV-based therapy development for the treatment of IBD.

bioinformatics↗

bioRxiv2020

ViralLink: An integrated workflow to investigate the effect of SARS-CoV-2 on intracellular signalling and regulatory pathways

The SARS-CoV-2 pandemic of 2020 has mobilised scientists around the globe to research all aspects of the coronavirus virus and its infection. For fruitful and rapid investigation of viral pathomechanisms, a collaborative and interdisciplinary approach is required. Therefore, we have developed ViralLink: a systems biology workflow which reconstructs and analyses networks representing the effect of viruses on intracellular signalling. These networks trace the flow of signal from intracellular viral proteins through their human binding proteins and downstream signalling pathways, ending with transcription factors regulating genes differentially expressed upon viral exposure. In this way, the workflow provides a mechanistic insight from previously identified knowledge of virally infected cells. By default, the workflow is set up to analyse the intracellular effects of SARS-CoV-2, requiring only transcriptomics counts data as input from the user: thus, encouraging and enabling rapid multidisciplinary research. However, the wide-ranging applicability and modularity of the workflow facilitates customisation of viral context, a priori interactions and analysis methods. Through a case study of SARS-CoV-2 infected bronchial/tracheal epithelial cells, we evidence the functionality of the workflow and its ability to identify key pathways and proteins in the cellular response to infection. The application of ViralLink to different viral infections in a cell-type specific manner using different available transcriptomics datasets will uncover key mechanisms in viral pathogenesis. The workflow is available on GitHub (https://github.com/korcsmarosgroup/ViralLink) in an easily accessible Python wrapper script, or as customisable modular R and Python scripts.Author summary Collaborative and multidisciplinary science provides increased value for experimental datasets and speeds the process of discovery. Such ways of working are especially important at present due to the urgency of the SARS-CoV-2 pandemic. Here, we present a systems biology workflow which models the effect of viral proteins on the infected host cell, to aid collaborative and multidisciplinary research. Through integration of gene expression datasets with context-specific and context-agnostic molecular interaction datasets, the workflow can be easily applied to different datasets as they are made available. Application to diverse SARS-CoV-2 datasets will increase our understanding of the mechanistic details of the infection at a cell type specific level, aid drug target discovery and help explain the variety of clinical manifestations of the infection.Competing Interest StatementThe authors have declared no competing interest.View Full Text

systems biology↗

bioRxiv2020

Antibiotic-induced disturbances of the gut microbiota result in accelerated breast tumour growth via a mast cell-dependent pathway

The diverse community of commensal microbes that comprise the gut microbiota is known to play an integral role in human health, not least through its ability to regulate host immune responses and metabolic pathways. Alterations to the homeostasis of this community, including through the use of broad-spectrum antibiotics, have already been associated with the progression of several cancers, namely melanoma and liver. The aggressive nature of breast cancer (BrCa), largely due to its ability to metastasize early, has ranked the disease with the second highest mortality rate of all cancers globally. Yet the body of research into the complex relationship between the microbiota and BrCa is still limited. This study found that a depletion of the microbiota, through the administration of antibiotics, significantly increased the rate of primary tumour progression in mouse BrCa models. We show that antibiotic-induced microbiota disturbances lead to changes in behaviour of a relatively obscure tumour-immune cell population: mast cells. We observed increases in tumour stroma-associated mast cells in antibiotic treated animals. Moreover, inhibition of mast cell degranulation, via cromolyn, slowed tumour progression in antibiotic treated animals but not in control animals. Thus, it appears that a perturbed microbiota drives stroma-associated mast cell recruitment and activation, which in turn promotes primary tumour growth through an as yet unknown mechanism. One Sentence Summary: We show that breast cancer progression is accelerated through a unique/novel immune response involving mast cells as a result of an antibiotic induced perturbation of the gut microbiota in a mouse model.

cancer biology↗

bioRxiv2019

A systems genomics approach to uncover patient-specific pathogenic pathways in a complex disease

We describe a novel precision medicine workflow, the integrated single nucleotide polymorphism network platform (iSNP), designed to identify the exact mechanisms of how SNPs affect cellular regulatory networks, and how SNP co-occurrences contribute to disease pathogenesis in ulcerative colitis (UC). Using SNP profiles of 377 UC patients, we mapped the regulatory effects of the SNPs to a human signalling network containing protein-protein, miRNA-mRNA and transcription factor binding interactions. Unsupervised clustering algorithms grouped these patient-specific networks into four distinct clusters based on two large disease hubs, NFKB1 and PKCB. Pathway analysis identified the epigenetic modification as common and the T-cell specific responses as differing signalling pathways in the clusters. By integrating individual transcriptomes in active and quiescent disease setting to the patient networks, we validated the impact of non-coding SNPs. The iSNP approach identified regulatory effects of disease-associated non-coding SNPs, and identified how pathogenesis pathways are activated via different genetic modifications.

bioinformatics↗