Multi-target drug design has emerged as a promising strategy to overcome the limitations of single-target drugs, particularly in combating infectious diseases that develop resistance through different mechanisms. Enzymes in the Menaquinone (MK) biosynthesis pathway of Staphylococcus aureus represent attractive targets for this approach, as they catalyze the sequential steps in the pathway, each of which processes a substrate intermediate that contains a thioester-coenzyme A or dicarboxylate moiety. Therefore, this acts as a vertical targeting strategy where multiple enzymes within the same pathway can be inhibited simultaneously. In this study, we provide the first systematic approach for vertical multi-target design. We used an in silico fragment-based approach to match the structural features of the active sites and identify common pharmacophoric elements, and screen for compounds that can simultaneously inhibit target enzymes within the MK pathway. Common fragments across the targets were identified through multi-target docking, and built compounds were prioritized using a geometric mean-based multi-target activity score. Structure-based pharmacophore modeling revealed a conserved Donor-Hydrophobic-Ring-Ring motif across all active sites. Molecular dynamics simulations supported the stability and multi-target potential of the identified compounds.
Heart failure with preserved ejection fraction (HFpEF) involves interacting immune, vascular, and stromal abnormalities. We asked whether macrophage genes showing opposite diet-associated and dapagliflozin-associated effects could identify regulatory and signaling programs relevant to HFpEF. Genes were ranked in a 2 × 2 × 2 dataset of sorted murine cardiac macrophages, and the locked signatures were evaluated in cardiac single-cell and single-nucleus datasets. Regulon analysis and network-constrained sensitivity testing identified BHLHE40 as the more robust program-level candidate. NFKB1, in turn, was linked to the broadest curated communication network, including TNF-, IL1B-, and PDGFB-related endothelial and fibroblast branches. External evidence varied across datasets and cell compartments. In myocardium from 19 HFpEF and 24 control donors, the CCR2- and cross-subset signatures were higher in macrophages, and the CCR2- signature was also higher in fibroblasts. PDGFB-related fibroblast branches received broader external evidence than the discovery-ranked TNF-TNFRSF1A endothelial branch, whose direction was not retained in human myocardium. These results nominate testable macrophage regulatory and signaling hypotheses for HFpEF but do not establish drug-specific reversal or cross-model conservation.
Multiple sclerosis (MS) is a chronic inflammatory disease characterized by demyelinating lesions in the central nervous system. While animal models have provided insights into lesion development, human-based studies remain limited. This study investigated the molecular landscape of gray matter lesions at different stages using spatial transcriptomics. Post-mortem cortical tissue from MS patients was analyzed to examine gene expression in lesions classified as early active, chronic active, or chronic inactive based on myelin integrity and macrophage/microglia activity. Distinct transcriptomic profiles were observed across lesion stages, although these findings should be considered exploratory given the limited sample size, with the early active category represented by a single lesion. The early active lesion showed increased expression of immune-related genes (e.g., HLA-DRA, CD68, CCL5) and enrichment of inflammatory and extracellular matrix remodeling pathways. Chronic active lesions showed enrichment of innate immune pathways, including Toll-like receptor and complement signaling, whereas chronic inactive lesions exhibited increased expression of heat shock proteins (HSPA1A, HSPA1B), and stress-response pathways. Control samples showed preserved neuronal and myelin gene expression. These findings provide preliminary insights into molecular processes that may be associated with MS lesion progression and may identify candidate biomarkers for future validation studies.
The gut microbiome shapes systemic physiology through metabolites that enter circulation, yet most computational approaches focus on predicting metabolite profiles from microbial features rather than inferring microbial composition from host metabolomes. Here, we investigate whether host-derived metabolomic profiles can be leveraged to predict gut microbial community structure and to determine how disease-associated dysbiosis reshapes metabolite-microbe interactions and gut-to-systemic metabolic communication. We developed an integrative multi-omics framework combining serum and cecal metabolomics with 16S rRNA-based microbiome profiling. Supervised learning models demonstrated that cecal metabolites carry predictive signals for microbial abundances across conditions. Regularized canonical correlation analysis (rCCA) revealed cross-compartment metabolite-microbe networks. These analyses showed both conserved and condition-specific interaction patterns, indicating substantial network reorganization under disease-associated dysbiosis. Pathway-level integration further identified metabolic pathways linking the gut microbiome, the cecal environment, and the systemic circulation, representing coordinated gut-to-systemic communication axes. Together, our results establish a multi-omics strategy for predictive inference of gut microbial composition from host metabolomes and provide a framework for identifying pathway-level mechanisms underlying host-microbe metabolic crosstalk.
Inflammatory bowel disease (IBD) involves complex immunometabolic dysregulation. Mitochondria-associated endoplasmic reticulum membranes (MAMs) link metabolic adaptation and inflammatory signaling, but gene expression features associated with MAM-related transcriptomic states in IBD remain unclear. Public transcriptomic datasets were integrated with a curated MAM-related gene set and herbal target information. Differential expression analysis, weighted gene coexpression network analysis, and machine learning-assisted feature selection were used to identify candidate genes. A two-gene nomogram was constructed and evaluated in discovery and independent validation datasets. Functional enrichment, immune deconvolution, regulatory network prediction, molecular docking, 200-ns molecular dynamics simulations, and preliminary quantitative polymerase chain reaction validation were performed. LIMK1 and PKM were identified as candidate transcriptomic signature genes associated with IBD status and MAM-related transcriptional states, showing discriminatory performance in independent datasets. Both genes were associated with cytokine-cytokine receptor interaction and drug metabolism-cytochrome P450 pathways and were linked to macrophage-related immune states. Structural analyses suggested stable predicted binding between PKM and 1-(4-hydroxybenzyl)-4-methoxy-9,10-dihydrophenanthrene-2,7-diol and structurally plausible interactions between LIMK1 and palmatine. Preliminary tissue-level validation supported the disease-associated expression pattern of PKM, whereas LIMK1 showed a consistent but nonsignificant upward trend. LIMK1 and PKM may represent candidate transcriptomic signatures associated with MAM-related regulatory states in IBD. These findings should be interpreted as hypothesis-generating associations rather than evidence of MAM localization, functional causality, or therapeutic efficacy. Further protein-level and functional validation is required.
Lung cancer remains a leading cause of cancer-related mortality worldwide due to its extensive molecular heterogeneity, late-stage diagnosis, and therapeutic resistance. Advances in high-throughput omics technologies have enabled comprehensive characterization of tumors across multiple biological layers, including genomics, epigenomics, transcriptomics, proteomics, and metabolomics. However, single-omics analyses provide only fragmented insights into tumor biology, highlighting the need for integrative multiomics approaches. Artificial intelligence (AI), particularly machine learning and deep learning, has emerged as a powerful tool for integrating heterogeneous datasets and uncovering biologically and clinically relevant patterns. This review summarizes recent advances in AI-driven multiomics integration for lung cancer, highlighting its applications in molecular subtyping, biomarker discovery, prognosis prediction, therapeutic response modeling, and precision oncology. We also discuss current challenges, including data heterogeneity, model interpretability, reproducibility, and clinical translation, together with emerging strategies for integrating multimodal data such as radiomics and digital pathology. Finally, we introduce precision phenomics as a unifying framework that links molecular, spatial, functional, and clinical characteristics of tumors to support personalized cancer management. Collectively, AI-driven multiomics integration has the potential to transform lung cancer research and improve patient outcomes.
The discussion on precision oncology integrates multiomics technologies and artificial intelligence, specifically addressing biomarker discovery and personalized therapeutic strategies. In this way, clinical translation and multiomics biomarkers are reconstructed challenges such as heterogeneity, validate, algorithmic bias, regulatory complexities, and ethical issues. This review critically evaluates how advanced technologies operating in an integrated manner facilitate precision oncology by supporting fields such as genomics, transcriptomics, proteomics, metabolomics, and radiomics. To conduct the study, we performed literature search using standard databases such as PubMed, Web of Science, and Scopus, focusing on papers published between 2020 and 2025. We address the all aspects of biomarker identification and clinical applications; we employed a five-stage framework comprising multiparametric data generation, integration, biomarker discovery, rigorous validation, and regulatory implementation. To further examine this review, we have employed emerging computational approaches, including machine learning and deep learning and graph neural networks alongside regulatory frameworks and ethical, legal, and social considerations. Discussing translation barriers, we consider factors such as limited reproducibility, validation, and critical discussion particularly studies. Ultimately, a future model based on standardized, validated, and transparent learning strategies accelerates and fosters the development of clinically reliable standards. Our review provides an integrative roadmap for modern, multiomics-driven biomarker approaches in precision oncology practice.
Dual-specificity protein kinase CLK4 plays a pivotal role in regulating alternative mRNA splicing, DNA repair, and various cellular processes through precise phosphorylation. In this study, we analyzed 3825 global human cellular phosphoproteome studies, identifying 430 qualitative profiles and 55 quantitative differential datasets featuring high-confidence Class-1 phosphosites (localization probability ≥ 75%; A-score ≥ 13). Notably, S136 and S138 emerged as predominant phosphorylation sites outside the kinase domain. These sites showed frequent detection and differential expression in liver, lung, and head and neck cancers, as documented in PhosphositePlus. We identified a high-confidence set of co-regulated phosphoproteins, including SQSTM1, SRRT, RPS6, TP53BP1, TNKS1BP1, THUMPD1, OTUD4, and TCEA1. These proteins link CLK4 to critical pathways, including RNA splicing, autophagy, DNA damage response, and cancer progression. Binary interactors, including SRRM2, Interacts with SPT6 1 (IWS1), RBBP6, ZC3H18, BUD13, and DYRK1A, further connect CLK4 to RNA processing and splicing. Predicted downstream substrates, such as CCNL2, CDK11B, PRDX6, YTHDC1, RBM15, SRRM1, SFSWAP, HNRNPU, and DOCK7, highlight CLK4's broad regulatory scope. Upstream kinases were also predicted for S136 and S138. Site-resolved analyses revealed tumor-specific dysregulation of CLK4 phosphorylation at key residues. Co-occurring phosphosite alterations and nearby somatic mutations suggest disrupted CLK4 regulation in cancer. Overall, this study provides a comprehensive phosphoproteomic resource that maps CLK4's co-regulatory networks, paving the way for mechanistic investigations and targeted cancer therapies.
This study investigated the connection between the transcriptome and proteome by integrating eQTL (Expression Quantitative Trait Locus) and pQTL (Protein Quantitative Trait Locus) datasets generated using different technologies. eQTL data were obtained from the eQTLGen (microarray-based) and INTERVAL (RNA-Seq-based) studies, while pQTL data were derived from the UK Biobank (Olink platform) and deCODE (SomaScan platform) studies. A total of 1162 genes common to all four datasets were analyzed. Mendelian randomization (MR) identified 211 genes whose transcript levels significantly (p < 5e-8) predicted protein levels, whereas genetic correlation analysis detected 67 genes with shared genetic regulation. Negative transcript-protein associations were observed for 12% of genes identified by MR and 7% of those identified through genetic correlation. Cross-platform comparisons showed the strongest concordance between eQTL and pQTL effect sizes in the INTERVAL-UK Biobank panel and the weakest in the eQTLGen-deCODE panel. Colocalization analysis further confirmed these findings and indicated genes with strong eQTL-pQTL overlap predominantly encode intracellular proteins, whereas genes with weak overlap tend to encode glycosylated secreted proteins. Integrating both the transcriptome and proteome for biomarker discovery and locus annotation is important, as the overall genetic architectures of the blood transcriptome and proteome are not the same. RNA-Seq and Olink platforms provide more accurate measurements of RNA and protein levels.
High-throughput shotgun proteomics is often hindered by the incompatibility of detergents with mass spectrometry (MS), making sample preparation a critical bottleneck for accuracy and robustness. This challenge is amplified in muscle proteomics, where the high dynamic range of protein abundance requires highly efficient and reproducible workflows to capture low-abundance proteins. To address this, we performed a systematic benchmarking of five preparation methods—stacking-gel (SG), tube-gel (TG), solid-phase extraction (SPE), filter-aided sample preparation (FASP), and suspension traps (S-TRAP)—using the sarcoplasmic fraction of pig muscle. The protein extracts obtained were subjected to label-free semi-quantitative proteomic analysis using high-performance nano-liquid chromatography coupled to tandem MS. Qualitative and quantitative results were compared using bioinformatics and biostatistics tools. Our study identified 530 proteins with significant variations across methods. S-TRAP provided the highest identification depth, capturing the broadest proteome coverage. Conversely, the TG method demonstrated superior quantitative reproducibility, essential for detecting subtle physiological changes. For translational research, such as meat quality science or muscle-related clinical models, our findings imply that S-TRAP is the preferred choice for discovery-phase proteomics (biomarker hunting), while TG or SG should be prioritized for high-precision validation studies.
Primary ovarian failure (POF) is linked to diabetes-related metabolic dysregulation, including inflammation, oxidative stress, and mitochondrial dysfunction. Summary-data-based Mendelian Randomization and colocalization analysis were employed to explore causal relationships between hypoglycemic drug targets and POF risk, integrating multi-omics data to uncover underlying genetic and metabolic mechanisms. A significant association was revealed between elevated dipeptidyl peptidase-IV (DPP4) expression levels and reduced POF risk. This association remained robust following multiple testing correction and colocalization analysis. In subsequent methylation level analysis, three CpG sites in DPP4 were identified, where elevated methylation levels were associated with increased POF risk. Furthermore, increased DPP4 protein levels were demonstrated to be associated with reduced POF risk. Through the integration of multi-omics evidence and two-sample Mendelian randomization analysis, the findings were further validated, with sensitivity analyses confirming the stability of the results and the absence of significant pleiotropy or heterogeneity. It was demonstrated that increased DPP4 gene expression and protein levels have protective effects against POF, whereas elevated methylation at specific CpG sites is associated with increased POF risk. Evidence is provided supporting DPP4 as a potential therapeutic target for POF prevention.
Fibromyalgia is a chronic pain syndrome characterized by widespread musculoskeletal pain, fatigue, sleep disturbances, and cognitive dysfunction, with substantial impact on quality of life and functional capacity. Despite its high prevalence, its underlying molecular mechanisms remain incompletely understood, and reliable biomarkers are lacking. This study performed an integrative analysis of publicly available transcriptomic datasets combined with protein-protein interaction network analysis, hub gene identification, and investigation of microRNA (miRNA)-mediated posttranscriptional regulation, in addition to evaluating molecular modulation following therapeutic intervention. Differentially expressed genes consistently identified across independent cohorts revealed two major molecular axes: An inflammatory-immune axis involving cytokine and interferon-related pathways, including IL6, TNF, CXCL8, and STAT1, and a structural axis associated with extracellular matrix organization, including COL1A1, COL3A1, FN1, and ITGB1. Integration with miRNA data demonstrated reduced expression of regulatory miRNAs linked to inflammatory and structural pathways, suggesting impaired posttranscriptional control. Therapeutic modulation analysis further demonstrated reduced expression of inflammatory genes and increased expression of structural genes following manual therapy, indicating partial reversibility of these molecular alterations. Collectively, these findings support the presence of multilevel molecular dysregulation in fibromyalgia and highlight potential biomarkers and therapeutic targets associated with inflammatory and peripheral structural mechanisms.
Metabolomics is a powerful systems-level approach and has the potential to serve as an important part for understanding the biochemical pathways and metabolic phenotypes in physiological and pathological states. Extracellular vesicles (EVs), which include apoptotic bodies, microvesicles, and exosomes, have emerged alongside metabolomics as active mediators of intercellular communication, transporting diverse cargo that mirrors the cellular origin and metabolic status of their parent cell. EVs are enriched with lipids, proteins, nucleic acids, and biologically active metabolites involve in signal transduction, metabolic regulation, and pathogenic mechanisms of a disease. Nevertheless issues related to heterogeneity of vesicles, purity of isolation, and detection sensitivity of metabolites, the study of EV metabolomics still methodologically and analytically challenging. This review offers a critical synthesis of current knowledge in EV metabolomics including analytical technology, statistical and computational approaches, and emerging clinical applications. In addition, a specific focus on methodological variability, contamination chances, and limitations in existing research study design that affect reproducibility and translation. Furthermore, multi-omics integration and machine learning are reviewed as promising approaches to enhance the discovery of biomarkers and interpretation of the biological system. Finally, highlighting the key research gap and future research directions to steer the advancement of clinically related applications in translational medicine.
Geroscience offers a transformative paradigm by targeting shared aging hallmarks to enable simultaneous modulation of multiple age-related disorders (ARDs). Yet, current geroprotective interventions often lack mechanistic breadth, as targeting isolated pathways yields limited benefits compared to interventions modulating interconnected regulators of aging biology. To bridge this gap, a systems-level strategy was designed around four key targets, including, Nrf2/Keap1, mTORC1, AMPK, and SIRT1, responsible for regulating oxidative stress, mitochondrial dysfunction, proteostasis, and autophagy. Concurrent regulation of these targets was identified to potentially induce a concerted and sustained geroprotective effect across diverse ARDs. A machine learning-based geroprotector classification model was developed to identify natural compounds capable of executing this integrated strategy. Subsequent drug-likeness screening confirmed favorable pharmacokinetic properties of the predicted compounds, while molecular docking revealed compounds with strong binding affinities with all four geroprotective targets, thereby leading to the identification of a subset of natural compounds with the potential to induce a coordinated geroprotective response. Finally, a graph neural network-based synergy prediction model, trained on known ARD drug combinations, identified five high-confidence pairs composed of four natural compounds, including Baicalein, Pectolinarigenin, Phloretin, and Demethoxycurcumin. These computationally predicted combinations hold the potential to elicit synergistic and comprehensive geroprotective effects across multiple ARDs.
Intervertebral disc degeneration (IVDD) is a major contributor to chronic low back pain and involves extracellular matrix degradation, inflammation, oxidative stress, and immune cell infiltration. This study integrated transcriptomic analysis, network pharmacology, molecular docking, molecular dynamics (MD) simulation, and ADMET prediction to explore potential mechanisms of Coptidis rhizoma (CR) in IVDD. Bioactive CR compounds and predicted targets were identified from TCMSP, SwissADME, and SwissTargetPrediction, and IVDD-related differentially expressed genes were obtained from two GEO datasets after normalization and batch-effect correction. Thirty-six overlapping genes were identified, and network analysis prioritized CCR1, CXCR2, ICAM1, TNF, STAT3, MPO, and MMP9 as core targets. Enrichment analyses highlighted chemokine signaling, cytokine-cytokine receptor interaction, leukocyte transendothelial migration, TNF signaling, IL-17 signaling, and NF-κB signaling. Docking and benchmark analyses indicated favorable predicted binding of Obacunone-CCR1 and Quercetin-MMP9, while 100-ns MD simulation supported the stability of the Obacunone-CCR1 complex. These findings suggest that CR may modulate chemokine-driven immune recruitment and downstream inflammatory-catabolic processes in IVDD, providing a hypothesis-generating basis for experimental validation.
Ectonucleoside triphosphate diphosphohydrolase 2 (ENTPD2), an enzyme involved in extracellular nucleotide metabolism and purinergic signaling, has been linked to tumor-immune interactions, although its role in colorectal cancer (CRC) remains unclear. This study examined the expression pattern and regulatory context of ENTPD2 through integrative analysis of transcriptomic, proteomic, microRNA (miRNA), and single-cell transcriptomic datasets. Transcriptomic analyses showed that ENTPD2 mRNA levels are elevated in colorectal tumors compared with normal tissues and that higher expression is associated with shorter relapse-free survival. In contrast, proteomic analyses indicated reduced ENTPD2 protein abundance in tumor samples, suggesting a divergence between transcript and protein expression. Analysis of candidate miRNAs identified miR-708-5p as a potential post-transcriptional regulator, supported by its increased expression in CRC and a predicted binding site within the ENTPD2 3 '-untranslated region (UTR). Single-cell transcriptomic datasets further indicated that ENTPD2 transcripts are mainly detected in malignant epithelial cells. We performed a functional validation using dual-luciferase reporter assays, qRT-PCR, and Western blot analysis in CRC cell lines. Experimental analyses demonstrated that miR-708-5p directly targets the ENTPD2 3 ' UTR in HCT116 cells and suppresses ENTPD2 expression in both HCT116 and HT-29 cells. These findings support a potential contribution of miR-708-5p to ENTPD2 regulation in CRC.
NIMA-related kinase 4 (NEK4) is a serine/threonine kinase implicated in microtubule stabilization, cilia function, and DNA damage response (DDR), with emerging roles in cancer progression through context-dependent effects on proliferation, epithelial-to-mesenchymal transition (EMT), and metastasis. Despite its significance, site-specific phosphorylation dynamics of NEK4 remain underexplored. Here, we conducted a comprehensive computational phosphoproteomic analysis by curating Class-1 phosphosites from over 3800 public datasets, identifying NEK4 phosphosites, including four predominant sites (S563, S661, S461, S639) outside the kinase domain that exhibit high detection frequencies and differential regulation. Coregulation analysis revealed phosphosites in other proteins (PsOPs) that coordinate with these NEK4 sites, linking them to DDR pathways (e.g., via interactions with DNA-PK complex components), EMT signaling, microtubule organization, and mitochondrial function. Network mapping integrated predicted upstream kinases (e.g., CDK13, RPS6KA1/3), downstream substrates (e.g., MKI67, INCENP), and binary interactors (e.g., TMPO, RRP1B), highlighting NEK4's integration into cancer-associated networks involving cell cycle regulation, apoptosis, and autophagy. Functional enrichment underscored NEK4's potential in modulating genotoxic stress responses and tumorigenic reprogramming. These findings provide a phospho-centric framework for NEK4 signaling, positioning it as a therapeutic target in DDR-defective and EMT-driven cancers, and lay the groundwork for experimental validation of its site-specific roles.
Coronavirus disease 2019 and pulmonary arterial hypertension (PAH) are clinically distinct disorders that converge on severe pulmonary vascular dysfunction, endothelial injury, and cardiopulmonary failure. However, the shared systems-level molecular architecture linking acute, virus-induced vascular damage with chronic pulmonary vascular remodeling remains undefined. To address this critical gap, we conducted a novel, integrative multilayered network analysis. By simultaneously combining transcriptome profiles of lung tissue samples with protein-protein interaction, metabolic, and regulatory networks, we systematically compared the previously unmapped molecular landscapes of both conditions. Despite their distinct upstream triggers-immune receptor-dominated signaling in acute severe acute respiratory syndrome coronavirus 2 infection versus remodeling- and ion channel-associated signaling in PAH-cross-disease integration revealed a highly structured, convergent immunometabolic regulatory architecture. This core is defined by extensive transcriptional reprogramming and the coordinated rewiring of oxidative phosphorylation and acetyl-CoA-associated metabolism. Taken together, these findings define a systems-level convergence model that mechanistically links acute viral endothelial injury and chronic pulmonary vasculopathy through a shared immunometabolic axis. This integrative molecular framework provides a new foundation for prioritizing candidate biomarkers and therapeutic nodes that target the overlapping vascular and inflammatory mechanisms of both conditions.
Malaria caused by Plasmodium falciparum remains a health burden worldwide due to drug resistance and limited treatment options. Calcium-dependent protein kinase 1 (CDPK1) plays a central role in parasite development and invasion, but the downstream molecular alterations that occur upon its disruption remain poorly understood. We present a proteogenomic-based data analysis pipeline for the reanalysis of the publicly available P. falciparum CDPK1 mutant dataset (PRIDE: PXD005207), integrating proteomic and phosphoproteomic data with six-frame genome translation. This led to the discovery of 24 new protein-coding genes, including 17 exonic and 7 intronic peptides, thereby enriching the current genome annotation. Several peptides, such as NILLTFDK, THNNNPQPNPQQK, and EVTSNFGNIR, mapped to previously unannotated genomic regions, which showed orthologous evidence in other Plasmodium species. The reanalysis of phosphoproteomics data identified 37 novel peptides that imply changes in phosphorylation signaling upon CDPK1 knockdown. The identification of conserved peptides like those associated with metacaspase and HSP70, indicates their potential roles in the survival and adaptation of parasites. Overall, this study highlights the potential of proteogenomics to improve genome annotation and reveal hidden coding regions of the P. falciparum genome. This provides new insights into kinase-regulated pathways and potential molecular targets for malaria control.
Inflammatory bowel disease (IBD) is a chronic and recurrent gastrointestinal disease, the pathogenesis of which has not been fully elucidated. Increasing evidence suggests that the disorder of mitochondrial metabolism is closely related to the pathogenesis of IBD, but its specific regulatory network and key genes remain to be further investigated. IBD-related transcriptome datasets (GSE3365 and GSE75214) and single-cell sequencing dataset (GSE134809) were obtained from the Gene Expression Omnibus database. Differentially expressed genes and hub genes were identified through differential expression analysis and weighted gene co-expression network analysis, and candidate genes were obtained by intersecting these with mitochondrial metabolism-related genes, followed by functional enrichment analysis. Machine learning algorithms were used to screen key genes and construct risk prediction models. Additionally, analysis of GSE134809 single-cell data identified characteristic cell types and expression distribution of key genes in IBD and explored communication between different cell types. Furthermore, immune cell infiltration, competitive endogenous RNA (ceRNA) network, and transcription factor prediction were performed. Finally, the diagnostic performance of key genes was validated in GSE75214 and reverse transcription-quantitative polymerase chain reaction. Two key genes, mitochondrial ribosomal protein L35 (MRPL35) and MRPL39, were identified, which were downregulated in IBD, and had good diagnostic potential. Single-cell analysis revealed that key genes were predominantly highly expressed in mononuclear phagocyte (MNP) cells. MNP cells communicated with other cells through receptor ligands including MIF-(CD74 + CXCR4), MDK-SDC1, and ITGB2-ICAM2, which are complexly related to mitochondrial metabolism. With the progression of IBD, infiltration levels of resting natural killer cells, naive B cells, M2 macrophages, and naive CD4 T cells decreased, and correlations between different cells continuously changed. A ceRNA network centered on XIST, hsa-miR-103a-3p, and MRPL35 was constructed. Additionally, therapeutic drugs targeting key genes were predicted, including cimetidine, eugenol, chlortetracycline, vincristine, irinotecan, bisacodyl, and sulpiride, with molecular docking validating high affinity between these drugs and key targets. This study constructed a multiomics integrated analysis strategy and identified MRPL35 and MRPL39 as potential markers and therapeutic targets, providing new insights for the diagnosis and treatment of IBD.
