Type I interferon (IFN) signaling is one of the body’s earliest and most important defenses against viral infection, rapidly activating hundreds of antiviral genes. In the December 2025 Journal of Virology article, “Inhibition of type I interferon signaling is a conserved function of gamma-herpesvirus-encoded microRNAs”, Fachko and colleagues demonstrate that gamma-herpesviruses, closely related to Epstein–Barr virus and Kaposi’s sarcoma–associated herpesvirus, encode microRNAs that consistently inhibit this pathway. By combining reporter assays, primary cell infections, and genetically engineered viruses lacking specific microRNA clusters, the authors demonstrate that viral microRNAs reduce interferon-stimulated gene expression during early infection and make latently infected cells less responsive to interferon. Importantly, they identify direct targeting of interferon receptors (IFNAR1 and IFNAR2) and central JAK/STAT pathway components (including JAK1, IRF9, and STAT-associated transcriptional regulators), revealing a multi-level strategy by which these viruses suppress antiviral immunity.The researchers use Reactome as they reanalyze Argonaute PAR-CLIP datasets, identifying “Interferon signaling” and “Interferon alpha/beta signaling” pathway host genes bound and regulated by viral microRNAs within the immune response pathway. This pathway-based analysis allowed them to move beyond individual gene hits and show that viral microRNAs converge on multiple nodes of the same antiviral signaling network. The use of high-quality Reactome pathway data strengthened the mechanistic conclusions of the study, highlighting how pathway-level analysis can reveal conserved immune evasion strategies employed by herpesviruses across species.

Progesterone, a hormone cyclically produced during menstrual cycles and used in hormone replacement therapy (HRT) after menopause, can promote cell proliferation primarily through a paracrine signaling mechanism, where progesterone receptor (PR)-positive 'luminal mature' cells secrete signaling factors that act on neighboring PR-negative 'luminal progenitor' cells. This mechanism plays a crucial role in normal mammary gland development and has been implicated in breast cancer pathogenesis. Anti-progestin therapy has long been regarded as a potential strategy for breast cancer prevention. Simões et al. (2025) in their November 2025 Nature article, Anti-progestin therapy targets hallmarks of breast cancer risk, report findings from the single-arm phase II trial (BC-APPS1; NCT02408770), showing the effects of ulipristal acetate (UA), a progesterone receptor antagonist, on breast tissue from women at higher risk of breast cancer. The study combined contrast-enhanced magnetic resonance imaging (MRI) data with multi-OMICs and imaging analyses of paired vacuum-assisted breast biopsies collected before and after 12 weeks of daily ulipristal acetate (UA) treatment. UA treatment reduced epithelial proliferation and depleted luminal progenitor cells, impairing their colony-forming capacity. Multi-omics analysis, including single-cell transcriptomics and laser-capture microdissection (LCM) proteomics, identified the extracellular matrix as the primary target of UA. Pathway enrichment using Reactome data revealed that extracellular matrix (ECM) processes were downregulated in fibroblasts and basal-myoepithelial cells while luminal hormone-sensing cells (LHS) showed downregulation of “RNA-processing” components and upregulation of matrix metalloproteinases associated with “collagen degradation”.  Among the downregulated ECM genes, collagen VI chains (COL6A2, COL6A3) were the most significantly reduced. CellChat and NicheNet analyses showed that UA reduced fibroblast and basal–myoepithelial collagen-signaling outputs and linked LHS-derived ligands (WNT5A, RARRES1, APOD) to the regulation of key collagen genes (COL6A3, COL1A2) in fibroblast subclusters, highlighting progesterone-dependent paracrine control of the breast matrisome. Imaging analyses confirmed decreased abundance of collagen I, collagen VI (COL6A3), and fibronectin (FN1)  which correlated with reduced tissue stiffness and MRI-assessed fibroglandular volume following UA treatment. Primary human breast organoids grown in stiff hydrogels showed increased expression of PR target gene TNFSF11 and luminal progenitor markers SOX9 and KIT, along with increased mammosphere formation - effects fully suppressed by UA treatment. Collectively, these findings reveal how ulipristal acetate modulates both epithelial and stromal biology through coordinated suppression of luminal progenitors and increased ECM remodeling to reduce breast cancer risk in premenopausal women.

Neuroblastomas are driven by MYCN hyperactivity, accumulate abnormally high amounts of arginine, proline, and ornithine, and produce abnormally high amounts of polyamines due to MYCN upregulation of ornithine decarboxylase (ODC), the rate limiting enzyme of polyamine synthesis. Difluoromethylornithine (DFMO), an inhibitor of ODC, has therapeutic effect against neuroblastoma.  Therefore, Cherkaoui et al (2025) in their October, 2025 Nature article "Reprogramming neuroblastoma by diet-enhanced polyamine depletion" tested whether a diet free of proline and arginine, precursors of ornithine in neuroblastoma, would improve survival further. Although the proline-arginine-free diet alone had no effect on survival, in combination with DFMO it approximately doubled survival in experimental models.

Cherkaoui et al. then investigated the mechanism by which DFMO combined with depletion of proline and arginine inhibited growth of neuroblastomas. The treated tumors exhibited a ten-fold reduction in polyamine content relative to untreated tumors. Because spermidine and other polyamines can enhance translation, the translation efficiency of genes was measured by large scale analysis of RNA (RNA-seq), ribosome-bound RNA (Ribo-seq), and proteins. The surprising finding was that in conditions of low polyamines, ribosomes stalled more frequently at codons with adenosine at the third position. This may be due to the combination of a requirement for highly modified tRNAs to translate these codons, and lower hypusination of the eIF5A translation factor.

Why would stalling at a particular set of codons produce a specific anti-cancer effect? Cherkaoui et al. employed the Reactome database to examine the codon usage in the genes encoding components of biological pathways. Surprisingly, the genes of the cell cycle pathway had a significantly higher proportion of codons ending in adenosine than did the genes of the neuronal system pathway, accounting for the selective effect of DFMO and the arginine-proline-free diet on neuroblastoma cell proliferation. Also surprisingly, all pathways varied significantly in codon usage, suggesting possible new methods of therapeutically regulating them.

In their September 2025 Nature Biomedical Engineering paper, An immune-competent lung-on-a-chip for modelling the human severe influenza infection response, Ringquist et al. show the importance of including tissue-resident and circulating immune cells, as well as stromal cells, in lung organoid chips to obtain a more realistic human in vitro model system for studying viral respiratory infections. Human-centric Reactome pathway enrichment analysis employed in this study shows that genes expressed in immune and stromal cells, mediating immune response and extracellular matrix remodeling, respectively, are among the top 10% upregulated in influenza H1N1-infected lung organoids, amid a global transcriptional shutdown, showing the value this ex vivo model for studying lung infectious disease.

In the February 2025 issue of Nature Communications, Carli et al. reported the development of a predictive model of cell line drug sensitivity from RNA-seq data using machine learning approaches. The model leveraged Reactome pathways in combination with large language models (LLMs) to provide a mechanistic foundation. It demonstrated strong performance and was applied to predict patient drug responses, with predictions supported by experimental validation. This work highlights how Reactome provides a robust framework for enhancing the interpretability of machine learning models in precision medicine.

In the May 2025 issue of Science Advances,  Dirvin et al. used single-cell transcriptomics and network-based algorithms on primary human airway cells to identify the key master regulator proteins hijacked by SARS-CoV-2, and then performed a large-scale screen to find drugs capable of reversing these effects. They used Reactome pathway analysis to characterize the biological processes, including membrane trafficking and  infectious disease pathways,  that were modulated by the eleven most promising drug candidates.

[July 1, 2025] Gene expression levels in normal and diseased human tissues show circadian variation, but studying this variation directly is difficult. In their May, 2025 PLoS paper, Ananthasubramaniam and Venkataramanan applied unsupervised machine learning to high-throughput omics data from primary human adenocarcinomas to identify circadian expression rhythms in hundreds of genes. Reactome gene set enrichment analysis identified genes with rhythmic expression patterns in multiple tumor types, significantly overrepresented in pathways of mitochondrial translation, respiratory electron transport, mitotic cell cycle, and adaptive immune system. The rhythmic expression of gene / protein targets of many FDA-approved and potential anti-cancer drugs in the adenocarcinomas suggests that timing of anti-tumor drug administration may improve efficacy.

[June 1, 2025] The lack of interpretability in deep neural networks is a challenging issue in biomedical applications. In their 2023 Nature Communications study, “Interpreting biologically informed neural networks for enhanced proteomic biomarker discovery and pathway analysis” Hartman et al. used Reactome’s pathway hierarchical tree directly to develop multi-layered, biologically informed neural networks (BINNs) to address this issue and enhance proteomic biomarker discovery and pathway analysis. Reactome provided critical information on biological entity relationships, enabling the creation of BINNs that achieved high predictive accuracy in the identification of disease-relevant biomarkers and pathways in septic acute kidney injury and COVID-19 datasets. These BINNs outperformed traditional methods and provided experimentally testable molecular mechanistic explanations.

Reactome is committed to maintaining the highest standards of scientific accuracy. To help prevent the circulation of retracted research, we conduct regular, systematic reviews of all literature-backed assertions in our database. If a publication listed in the Retraction Watch database has been used as supporting evidence for any Reactome annotation, we re-evaluate the associated data. Annotations linked to retracted papers are either updated with new, valid references or removed entirely if no suitable replacements can be found. Each removal is documented along with the reason for the change. To date, we have reviewed over 40,000 curator-selected references and identified just 70 retracted articles. All affected annotations have been reviewed and revised accordingly. This ongoing process ensures the continued integrity and reliability of Reactome’s content.

In the March 2025 issue of Science Advances Yi et al. reported the development of a brain age estimation model using large-scale genetic and imaging data. Brain age gap (BAG) is a digital phenotype that may reflect associations with various brain disorders. This study aimed to identify potential drug targets causally associated with BAG. A total of 64 genes were identified within five Reactome pathways: programmed cell death, platelet signaling and aggregation , extracellular matrix organization, cell surface interactions at the vascular wall, and apoptosis Of these, seven genes were prioritized as targets due to strong genetic evidence for BAG.

In the January 2025 issue of Frontiers in Genetics,  Jensen et al. explored the relationship between DNA methylation patterns and adolescent brain development. By analyzing a cohort of adolescents aged 9 to 14, they identified co-methylation networks linked to cognitive improvements and structural brain changes. Pathway analysis using Reactome revealed that these networks are enriched in neuronal-related pathways, suggesting that epigenetic modifications play a significant role in the maturation of the adolescent brain.

In the July 2024 issue of Briefings in Bioinformatics, Yi et al.  report on The Biochemical Pathway Prediction (BPP) framework, a predictive analytical tool that utilizes various graph representation learning models to predict attributes and links in biochemical pathways. BPP provides two pieces of information: link prediction, which identifies potential connections between entities and reactions, and attribute prediction, which predicts missing attributes of nodes. The BPP framework was used to evaluate datasets derived from Reactome pathway data (version 75 to version 85), specifically identifying a key receptor, glycosylated-ACE2, instrumental in the SARS-CoV-2 viral invasion process.

In the October 2024 issue of Cell Reports Medicine, Tindle et al. identified two Crohn’s disease (CD) molecular subtypes - immune-deficient infectious CD (IDICD) and stress and senescence-induced fibrostenotic CD (S2FCD) - through multi-omics and functional analyses of patient-derived organoids. Reactome pathway enrichment analysis revealed subtype-specific dysregulations. In IDICD, the Nuclear receptor transcription factor pathway, Butyrophilin family interactions, and Intestinal infectious disease events were upregulated while Cytokine signaling in immune system events were downregulated. In S2FCD, Oncogene- and Oxidative stress-induced senescence pathways were upregulated and Signaling by TGF-beta receptor complex events were downregulated suggesting distinct subtype-specific therapeutic strategies.

In the August 2024 issue of PLoS One, Huang et al. used the Reactome database to analyze the pattern of RNA editing in cells in response to infection by SARS-CoV-2 and found that editing was highest in transcripts of genes related to immune response andcytokine production. Single cell transcriptomics showed that the Reactome Interferon signaling pathway is enriched in plasmacytoid B cells, B cells, and T cell subtypes during SARS-CoV-2 infection.

In PLOS Computational Biology, Wieder et al. (2024) employ the Reactome database and PathIntegrate, a pathway-based multi-omics integration tool based on single-sample pathway analysis and machine learning, to translate multi-omics datasets from molecular abundance measurements to pathway activity scores, enabling integration of disparate types of omics data according to a common scale. PathIntegrate provides higher sensitivity at low signal levels and efficiently identifies perturbed pathways from multi-omics datasets in COVID-19 and chronic obstructive pulmonary disease (COPD) examples, providing a readily interpretable predictive model.

In the feature article of the October 2024 issue of Drug Discovery Today titled “Chemical coverage of human biological pathways”, Kwak et al. describe the Target 2035 initiative, whose mission is to discover chemical tools for all human proteins by 2035. The authors use Reactome as the reference standard to determine the chemical coverage of human biological pathways and to outline the advantages of adopting the pathway-based rather than the proteome-based approach in guiding Target 2035 efforts.

In the August 2024 issue of Nature Microbiology, Bracha et al. use Reactome expression analysis to confirm that they successfully delivered multiple large (>100 kDa) therapeutic proteins across the blood-brain barrier into target neurons in mice using engineered Toxoplasma gondii secretion systems.

In their May 2024 paper in Cell Genomics, Arora et al. used a framework of Reactome signaling and metabolism pathways to integrate RHEA metabolic reactions and IUPhAR catalogs of G Protein-Coupled Receptors (GPCRs) and their ligands to define axes that combine signaling cascades and ligand metabolic processes. Altered expression of the sets of proteins that make up these axes correlate with patient survival cataloged in The Cancer Genome Atlas (TCGA) for many tumor types and suggest novel druggable targets.

In their May 2024 Nature Communications paper, Chen et al used data simulated based on Reactome pathways to validate their Functional Representation of Gene Signatures (FRoGS) algorithm, a deep learning-based approach that was designed to improve the accuracy of drug target predictions by addressing limitations of gene identity-based pathway analysis.

In the May 2024 issue of NPJ Precision Oncology, Cyberski et al. used Reactome’s hierarchically arranged pathways with their in silico Pathway Activation Network Decomposition Analysis (iPANDA) algorithm to identify upregulation of networks associated with cell cycle progression, signal transduction, and metabolism and down-regulation of immune cellular process and apoptosis in MYC-amplified cases of recurrent/metastatic head and neck squamous cell carcinoma.

In the May 2024 issue of Briefings in Bioinformatics, Xu et al. develop an Interpretable Biological Pathway Graph Neural Network (IBPGNET) framework based on Reactome pathway hierarchy to predict regulatory mechanisms that lead to lung adenocarcinoma recurrences. IBPGNET identified two genes of interest and performed in vitro knockdown models for drug sensitivity experimental validation. This study offers an approach for exploring molecular mechanisms underlying recurrence using Reactome’s hierarchical pathway structure.

In the December 2023 issue of Toxicology and Applied Pharmacology, Zhang et al. used the R package, ReactomePA (Yu and He, 2016), to identify enriched pathways responding to nickel-induced transcriptional memory changes in response to a second respiratory toxicant, nicotine. Nicotine exposure upregulated a specific subset of genes in the cells previously exposed to nickel, identifying a robust activation of Interferon (IFN) signaling, a driver of inflammation associated with many chronic lung diseases.

Using WGCNA, GO and KEGG data analysis tools, He Y et al. in the January 2024 issue of BMC Medical Genomics, established a connection among the genes and pathways associated with type 2 diabetes (T2D) and ischemic stroke (IS) and identified GRN (granulin precursor) as the hub gene in T2D-related stroke. The functional enrichment analysis using Reactome analysis tool for GRN identified Neutrophil degranulation, Toll-like Receptor Cascades, DDX58/IFIH1-mediated induction of interferon-alpha/ beta, and NLR signaling pathways as shared biological processes in T2D and IS.

In the July 2023 issue of BMB Reports, Lee et al. (2023) employed the Reactome pathway database and tools to predict the anti-cancer effects of rutin, a natural phytochemical identified as a lead chemotherapeutic against breast cancer by text mining Korean traditional medicinal compendia from 1596 CE and 1613 CE. Genes that may be associated with rutin's effects were analyzed for pathway enrichment and functional interactions by the Reactome Functional Interaction (FI) plugin app of Cytoscape. Focal adhesion and Apoptosis were among the pathways predicted to be affected by rutin and these effects were confirmed by treatment of breast cancer cells with rutin in vitro and in xenografts in mice.

In the November 2023 issue of Cell Communication and Signaling, Bukva et al. (2023) analyzed the proteomes of tumor-produced extracellular vesicles and identified sets of proteins that could discriminate tumor types, invasiveness, and proliferative capacity. In this analysis, 172 most predictive proteins were identified and used to classify nine tumor types with 91.67% efficiency. Reactome Pathway enrichment analysis of these proteins showed that each tumor type had perturbations in a distinct set of pathways. The proteins could be organized and used to discriminate the invasiveness and proliferative capacity of the tumors. High expression of proteins positively associated with invasiveness and proliferation correlated with reduced patient survival times.

In the chapter entitled “New Insights into Clinical Management for Sickle Cell Disease: Uncovering the Significant Pathways Affected by the Involvement of Sickle Cell Disease”, published in Methods in Molecular Biology 2024, Chouhan et al. describe the use of Reactome FIviz Cytoscape plugin to analyze pathway enrichment and construct a functional interaction network for DisGNET-derived sickle cell disease-associated genes, identifying genes involved in “Glucuronidation”, “Aspirin ADME”, “Phase II-Conjugation of compounds”, “Interleukin-4 and interleukin-13 signaling”, “Interleukin-10 signaling”, “Signaling by interleukins”, “Biological oxidations”, and “Cytokine signaling in immune system”.

In their paper titled “XMR: an explainable multimodal neural network for drug response prediction” published in Frontiers in Bioinformatics in August 2023, Wang et al. use five Reactome pathways, Cell Cycle, DNA repair, Disease, Signal transduction, and Metabolism, as an architecture of a visible neural network that is part of a deep learning model for prediction of drug responses in triple negative breast cancer.

In the  August 2, 2023 issue of Clinical Epigenetics, Miller et al. performed an epigenome-wide association study using Reactome  Functional Interaction network analysis and determined that DNA methylation at loci involved in calcium channel activity and development was associated with long-term cardiovascular disease risk beyond known risk factors in type 1 diabetes, particularly in individuals with greater glycemic exposure.

​​With current treatments, focal segmental glomerulosclerosis (FSGS), the largest cause of nephrotic syndrome, frequently progresses to end-stage kidney disease. Gebeshuber et al. (2023) assembled 376 FSGS-associated proteins into a FSGS pathophysiology model, major components of which were Reactome pathways for signal transduction and hemostasis. The 39 proteins shared between FSGS model and a 102-protein model for the antiplatelet drug clopidogrel included 20 therapeutic targets of the drug. Tested in an FSGS mouse model, clopidogrel significantly attenuated disease severity, repositioning the drug as an attractive candidate for human clinical trials for FSGS.

Using Reactome analysis tools and FIVIz, Balmorez et al. in the March 2023 issue of Int. J. Mol. Sci., established a commonality between the genes and pathways associated with Alzheimer's disease (AD), Ageing (AR) and Longevity. The pathways shared between AD and AR are p53-Dependent G1/S DNA damage checkpoint, FOXO-mediated transcription, and SUMOylation; between AD and longevity are Cytokine Signaling in Immune system, Plasma lipoprotein assembly, remodeling, and clearance, Metabolism of fat-soluble vitamins, and NR1H2- and NR1H3-mediated signalling; and between AR and Longevity are Immune system and Cytokine signaling.

The Zika virus (ZIKV) is an emergent arthropod-borne virus (arbovirus) responsible for congenital Zika syndrome (CZS) and a range of other congenital malformations. With little known about the pathways involved in CZS, Gratton et al in the February 2020 issue of Microorganisms conducted a meta-analysis of transcriptome studies to identify the genes and pathways altered during Zika infection. Reactome analysis identified interferon, pro-inflammatory, and chemokines signaling as well as apoptosis as key IFN signaling pathways in ZIKV-infected cells with three new candidate genes involved in hNPCs infection identified: APOL6, XAF1, and TNFRSF1.

Pregnancies complicated by Coronavirus Disease 2019 (COVID-19) are at an increased risk of severe morbidity. In multi-omics analyses investigating the pathophysiology behind severe COVID-19 disease, Altendahl et al, in the November 2022 issue of PLoS One found precipitous changes in maternal serum in those with severe COVID-19 infection. Reactome pathway enrichment analysis revealed upregulated analytes in 4 pathways: Complement cascade, Signaling by the B Cell Receptor (BCR), Fc epsilon receptor (FCERI) signaling, and FCGR activation.

Tagliazucchi L et al. 2023 in the March 2023 issue of ACS Infectious Diseases used the REACTOME overrepresentation and pathway topology analyses to identify Transport of small molecules, Cellular response to stress and other pathways associated with drug resistance and therapeutic failure (TF) during Leishmania infantum infection; they also discovered NDK3 and TFRC as potential targets for host-directed anti-Leishmania therapies to overcome drug-resistance.

The Reactome database helped Yu et al. in the January 2023 issue of the Journal of Experimental & Clinical Cancer Research identify candidate drugs for treatment of four subtypes of lung cancer that were categorized by pharmaco-genomic analysis of patient-derived cells and correlated with drug sensitivity of the patient-derived cells and with Reactome pathways identified by gene set variation analysis.

Analysis of all constituents of entire Reactome pathways identified by the presence of upregulated or mutated genes helped Valdivia et al. in the February, 2023 issue of Translational Vision Science & Technology to identify druggable targets and potential drugs for the treatment of diabetic retinopathy (DR). Drugs affecting MMP13 and LGALS3 in the regulation of myeloid cell differentiation by RUNX2 were notable candidates.

Reactome overrepresentation analyses of differentially expressed genes common to both Autism Spectrum Disorder and Tourette Syndrome help identify common targetable inflammatory pathways as described by Alshammeryet al. in the December 2022 issue of Frontiers in Neuroscience.

Gene set enrichment analysis (GSEA) conducted on Reactome’s carbohydrate metabolism pathways identifies the Pentose phosphate pathway and the Glucose metabolism pathway as the two most frequently upregulated pathways in tumors with high tumor-specific total mRNA expression (TmS) across 15 tumor types; TmS is a novel tumor phenotype-predictive quantitative feature described by Cao et al. in the November 2022 issue of Nature Biotechnology.

Reactome pathway gene sets in the MSigDB facilitated identification of the liver proteasome transcriptional switch that acts as the fasting timer in intermittent fasting in work published by Wei et al. in Cell Reports on October 25, 2022. The authors suggest that a 16-hour interval in intermittent fasting may be most beneficial for health.

Reactome pathway enrichment analysis helps to pinpoint expansion of regulatory T cells as a new biomarker of CAR T cell therapy resistance and toxicity in patients with B cell lymphoma. The study was published by Good et al. in Nature Medicine on September 12, 2022.