Machine learning identifies neutrophil-related signatures for diagnostic value in neonatal sepsis

Sepsis is a persistent systemic inflammatory disease involving multiple organ failure caused by a dysregulated immune response to infection. As primary effector cells in innate immunity, neutrophils significantly contribute to combating infections and mediating inflammatory responses. The aim of this study was to evaluate the prognostic significance of neutrophil-related genes (NRGs) in sepsis and their relationship with the immune microenvironment. This study was drawing on transcriptomic data and clinical characterization of sepsis patients from the GEO database. Sepsis prognostic genes were discovered through a combination of differential analysis, weighted gene co-expression network analysis (WGCNA), and univariate cox regression analysis. Molecular subtypes of sepsis were identified by consensus clustering methods, survival differences between subtypes were compared and gene enrichment analysis was performed. Further univariate cox regression analysis and LASSO combined were performed to identify sepsis feature genes. Utilizing multivariate cox regression analysis, a prognostic model was developed, and its predictive capability was subsequently examined through receiver operating characteristic (ROC) curve and Kaplan-Meier (K-M) survival analyses. Subsequently, immune cell infiltration and gene enrichment were assessed in the various risk groups. This study identified the presence of two molecular subtypes in sepsis, with Cluster1 patients having significantly better survival than Cluster2. The prognostic model based on NRGs (RCBTB2, KRT23, KLF9, GIMAP4 and CD84) in this study significantly differentiated the survival prognosis of high risk and low risk patients. The low risk patient group showed significant enrichment in immune related pathways (T cell receptor signaling and primary immunodeficiency), while the high risk group primarily exhibited significant enrichment in metabolism related pathways (glycine serine and threonine metabolism). Risk models constructed on the basis of NRGs are effective in predicting survival outcomes and immune profiles of sepsis patients, providing a new perspective on the link between NRGs and sepsis.

Chronic Rhinosinusitis with Nasal polyps (CRSwNP) are characterized by chronic inflammation and occur in 1-4% of the population worldwide. Patients often have comorbid asthma, and standard treatments among them are hindered by significant recurrence and lack of durability. Currently, knowledge of the molecular circuitry and immune microenvironmental interplay that utilizes metabolic reprogramming within CRSwNP is incomplete. Utilizing CRSwNP datasets from the GEO database, we performed bioinformatics analysis to identify differentially expressed genes (DEGs) implicated in metabolic reprogramming. Key regulatory genes were subsequently selected by weighted gene co-expression network analysis (WGCNA) and machine learning algorithms; their relationship with the immune microenvironment was then evaluated. To further investigate the underlying pathogenic mechanisms, we performed single-cell RNA sequencing (scRNA-seq) to map cellular expression patterns and applied Mendelian randomization (MR) analysis to assess potential causal relationships. Key molecules were subsequently experimentally validated by quantitative real-time PCR (qRT-PCR). We identified 21 DEGs associated with metabolic reprogramming that are relevant to CRSwNP. This subset was then analyzed using machine learning to identify 8 hub genes - ERBB4, FBP1, HMGCS2, LYZ, NDRG2, PIP, PYCR1, and SLC43A1. A prediction model built using these biomarkers yielded high diagnostic performance (AUC=0.979). Single-cell resolution analysis revealed that distinct expression patterns were exhibited by these genes across subsets of immune cells. MR analysis determined that lower expression of FBP1, LYZ and NDRG2 could be risk factors for CRSwNP. Subsequent qRT-PCR in independent samples validated the downregulation of these genes in CRSwNP tissues. We systematically identify and validate a set of metabolic reprogramming-related genes with diagnostic value in CRSwNP. Collectively, these findings not only heighten the current mechanistic understanding of CRSwNP pathogenesis but also offer a novel platform to devise diagnostic and therapeutic avenues focusing on metabolism.

The increasing prevalence of sarcopenia (SAR) has raised significant concerns in healthcare. Although mitochondrial dysfunction and immune disorders are recognized as risk factors, the interactions between them remain unclear. This study aims to identify potential diagnostic biomarkers associated with both phenotypes in the progression of SAR. Three transcriptional datasets were obtained from the GEO database. Gene set enrichment analysis (GSEA) was performed to explore the features of the training set, followed by filtering the differentially expressed genes (DEGs). Weighted gene co-expression network analysis was applied to select gene modules closely related to SAR. SAR-mitochondria-related DEGs were then determined by intersecting the DEGs, weighted gene co-expression network analysis, and mitochondrial-related genes from the MitoCarta3.0 database. Hub genes were further explored using LASSO and random forest machine learning algorithms. The regulatory molecules of these hub genes were predicted using the NetworkAnalyst database. Subsequently, receiver operating characteristic analysis was performed and the immune infiltration was analyzed using the CIBERSORT algorithm. A SAR model was established in C2C12 cells using d-galactose, and RT-qPCR experiments were performed for further validation. Gene set enrichment analysis results revealed that the training set genes are mainly enriched in mitochondrial function and energy metabolism. Through the machine learning methods, 5 hub genes were screened out from 32 SAR-Mito DEGs, namely MTRF1L, MICU1, DHTKD1, ACADM, and FHIT. A total of 42 transcription factors and 10 miRNAs strongly associated with the hub genes were detected. These hub genes demonstrated solid diagnostic potential in both training and validation sets. Furthermore, immune infiltration analysis indicated a significant reduction in neutrophil levels in SAR patients. The downregulation of the hub genes in d-galactose-induced C2C12 cells was confirmed. Collectively, our findings identified 5 potential biomarkers for the diagnosis and therapy of SAR and emphasized the interaction between mitochondrial function and the immune response in the development of this condition.

Identification and validation of genes associated with copper death in oral squamous cell carcinoma based on machine learning and weighted gene co-expression network analysis

Cancer‐associated fibroblasts (CAFs) are the center of cross‐communication between various cells in the tumor stroma. However, how CAFs‐associated genes play an important role in Head and neck squamous cell carcinoma (HNSCC) prognosis has not been reported. Transcriptome data were downloaded from TCGA and GEO databases. Devtools, DPIC, xCell, MCPcounter, and Estimate packages were used to calculate CAFs scores and immune infiltration. Prognosis and weighted gene coexpression network analysis (WGCNA) analysis were performed between high or low risk populations based on CAF scores. Hub genes were identified, intersected, and enriched between TCGA and GEO databases. CAFs related genes were used to construct a prognostic model and the tumor immune dysfunction and exclusion database was used to evaluate the immune infiltration. Drug sensitivity, difference analysis and the HPA database were used to identify sensitive drugs and verify their expression. TCGA and GEO data suggested that CAFs scores had a role in HNSCC prognosis prediction. Based on CAFs scores, WGCNA and core gene enrichment analysis were performed to construct a CAFs‐related prognostic model. The prognostic model composed of a total of 12 CAFs genes could predict the prognosis well and was validated in the validation dataset, demonstrating its applicability to external data. According to the model, although there was no statistical difference in immune escape between the high and low risk groups, the proportion of patients who responded to immunotherapy was different. Drug sensitivity also differed between the two groups. This study suggests that CAFs associated genetic signatures may help to optimize risk stratification and provide new insights into individualized cancer treatment.

Identification and initial experimental assessment of DCLK3 as a potential sarcopenia biomarker: A multi-omics and machine learning approach.

BackgroundDiabetic nephropathy (DN) is a common complication of diabetes, characterized by damage to renal tubules and glomeruli, leading to progressive renal dysfunction. The aim of our study is to explore the key role of metabolic reprogramming (MR) in the pathogenesis of DN.MethodsIn our study, three transcriptome datasets (GSE30528, GSE30529, and GSE96804) were sourced from the Gene Expression Omnibus (GEO) database. These datasets were integrated for batch effect correction and subsequently subjected to differential expression analysis to identify differentially expressed genes (DEGs) between DN and control samples. The identified DEGs were cross-referenced with genes associated with MR to derive MR associated differentially expressed genes (MRRDEGs). These MRRDEGs underwent Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. To identify key genes and develop diagnostic models, four machine learning algorithms were employed in conjunction with weighted gene co-expression network analysis (WGCNA) and the protein interaction tool CytoHubba. Gene set enrichment analysis (GSEA) and CIBERSORT analysis were conducted on the key genes to assess immune cell infiltration in DN. Additionally, a competitive endogenous RNA (ceRNA) network was constructed using the key genes. Finally, the expression levels of core genes in human samples were validated through quantitative real-time PCR (qRT-PCR).ResultsWe identified 256 MRRDEGs, highlighting metabolic and inflammatory pathways in DN. KEGG analysis linked these genes to the MAPK signaling pathway, suggesting its key role in DN. Six key genes were pinpointed using WGCNA, PPI, and machine learning, with their diagnostic value confirmed by ROC analysis. CIBERSORT revealed a strong link between these genes and immune cell infiltration, indicating the immune response’s role in DN. GSEA showed these genes’ involvement in inflammatory and metabolic processes. A ceRNA network was predicted to clarify gene regulation. qRT-PCR confirmed the expression patterns of CXCR2, NAMPT, and CUEDC2, aligning with bioinformatics results.ConclusionThrough bioinformatics analysis, a total of six potential MRRDEGs were identified, among which CUEDC2, NAMPT, CXCR2 could serve as potential biomarkers.

This study aimed to identify potential interacting genes between abdominal aortic aneurysm (AAA) and periodontitis. To achieve this, we obtained datasets of AAA and periodontitis from the GEO database, conducted differential analysis on the AAA dataset, and performed weighted gene co-expression network analysis (WGCNA) on the periodontitis dataset to preliminarily identify interacting genes via intersection. Subsequently, we refined key candidate genes by constructing a PPI network and applying three machine learning algorithms. These candidate genes were further validated through external independent datasets, receiver operating characteristic (ROC) curves, and Nomograms. Finally, single-gene Gene Set Enrichment Analysis (GSEA), immune landscape analysis, and targeted drug prediction were performed on the identified key genes. In our study, a total of 323 differentially expressed genes (DEGs) related to AAA and 4,412 periodontitis-related module genes were identified, producing 90 interacting genes through intersection initially. Through PPI network analysis and machine learning, we prioritized 7 key interacting genes. Validation confirmed that IL1B, PTGS2, and SELL were robustly associated with both diseases. Immune landscape assessment demonstrated that these three genes exhibited significant negative correlations with regulatory T cells (Tregs) and positive correlations with neutrophil infiltration. Additionally, ten drugs with the highest predicted target specificity were identified. In conclusion, we utilized various machine learning and bioinformatics approaches to preliminarily elucidate potential comorbid mechanisms between AAA and periodontitis from a multidisciplinary perspective.

Spinal cord injury (SCI) significantly impacts patients, with mitochondrial dysfunction playing a critical role in its pathology. Identifying mitochondria-related genes may offer new therapeutic and prognostic insights. RNA sequencing data from the GEO database were analyzed to identify differentially expressed genes (DEGs). Functional enrichment analyses were conducted, and weighted gene coexpression network analysis (WGCNA) alongside machine learning algorithms was used to identify key mitochondria-related genes. Immune infiltration was assessed using the EPIC algorithm, and single-cell RNA sequencing (scRNA-seq) data were analyzed for cellular diversity. A total of 2566 upregulated and 2634 downregulated genes were identified in SCI versus control samples. GO and KEGG enrichment analyses revealed these DEGs were primarily involved in oxidative stress, mitochondrial function, and immune pathways, including necroptosis and T cell receptor signaling. Then, 1578 genes with the strongest correlation to SCI were selected by WGCNA. By integrating DEGs, WGCNA module genes, and mitochondria-related genes, 76 candidate genes were obtained and used to construct a PPI network. Six hub genes (NDUFB3, SLC25A24, SLC25A40, GSTZ1, MAOA, and MRPL12) were identified by machine learning, all showing strong diagnostic potential (AUC > 0.77). Immune infiltration analysis indicated reduced B and T cell infiltration and increased macrophage activity in SCI samples. scRNA-seq analysis further revealed higher expression of NDUFB3 in dendritic cells and MAOA in pro-B cells, suggesting their involvement in immune regulation and mitochondrial dysfunction. These six genes represent potential biomarkers and therapeutic targets for SCI, providing insights into its molecular mechanisms and immune response.

Mitochondrial metabolic reprogramming in macrophages is crucial in the development and progression of inflammation. Given vitamin A's antioxidant properties and its therapeutic effects on inflammation, this study aims to elucidate how vitamin A influences mitochondrial metabolic reprogramming in inflammatory states, specifically in periodontitis, through genetic bioinformatics and experimental methods. The study utilized the GSE16134 dataset from the Gene Expression Omnibus (GEO) database, focusing on human periodontitis. Vitamin A-targeted genes (ATGs) were identified and analyzed using CIBERSORT to explore their role in inflammation. Cluster analysis revealed two phenotypes associated with ATGs, showing differential expression of genes like COX1, IL-1β, and STAT3, and immune activation patterns. Weighted Gene Co-expression Network Analysis (WGCNA) identified 145 markers correlated with ATG-guided phenotypes and inflammation. Machine learning models, combined with Gene Set Variation Analysis (GSVA), identified five key genes (RGS1, ACAT2, KDR, TUBB2A, TDO2) linked to periodontitis. Cell Type-Specific Enrichment Analysis (CSEA) highlighted macrophages as critical in metabolic reprogramming, validated by external datasets with an AUC of 0.856 in GSE10334 and 0.750 in GSE1730678. Experimental validation showed vitamin A's role in suppressing endoplasmic reticulum stress and altering mitochondrial dynamics, as well as metabolic reprogramming influencing inflammation via the STAT3 pathway in RAW 264.7 cells. The study identified 13 differentially expressed ATGs in periodontitis, showing strong correlations with inflammation, particularly in plasma cells, macrophages, dendritic cells, neutrophils, and mast cells. Two ATG-guided phenotypes were identified, differing in gene expression and immune activation. WGCNA and machine learning models identified 145 markers and five key genes associated with periodontitis. GSVA and CSEA analyses highlighted the JAK-STAT pathway and macrophage involvement in metabolic reprogramming. Experimental data confirmed vitamin A's effects on mitochondrial dynamics and metabolic reprogramming through the STAT3 pathway. The study demonstrates that vitamin A's therapeutic effect on periodontitis is mediated through JAK-STAT pathway-guided mitochondrial metabolic reprogramming in macrophages. It identifies two genetic and immune-related phenotypes and five genetic identifiers associated with periodontitis risk.

BackgroundOvarian cancer (OC) remains the most lethal gynecological malignancy, with profound tumor microenvironment (TME) heterogeneity contributing to the suboptimal clinical response to immune checkpoint inhibitors (ICIs). Therefore, delineating the metabolic heterogeneity of T cells within the TME is imperative for identifying precise prognostic biomarkers and refining individualized immunotherapeutic strategies.MethodsWe integrated single-cell RNA sequencing (scRNA-seq) data (GSE184880) with multi-omics profiles from The Cancer Genome Atlas Ovarian Cancer (TCGA-OV) cohort. The Seurat and Harmony algorithms were employed to identify tumor-specific T-cell subpopulations (CA-T cells). Weighted gene co-expression network analysis (WGCNA) and machine learning frameworks—including the least absolute shrinkage and selection operator (LASSO) regression and Random Survival Forest—were applied to screen core prognostic genes and construct a multi-gene risk scoring model (riskScore) for predicting overall survival (OS). The model was validated through Kaplan-Meier survival analysis, time-dependent receiver operating characteristic (ROC) curves, and independent prognostic assessment. Finally, ESTIMATE, CIBERSORT, and TIDE algorithms were applied to assess the associations between riskScore, TME immune landscapes, and ICI efficacy.ResultsWe identified a distinct tumor-specific T-cell subset (CA-T cells) significantly enriched in malignant tissues. This subset exhibited a functional signature characterized by mitochondrial metabolic reprogramming, specifically oxidative phosphorylation and electron transport chain activity. An 8-gene riskScore model—comprising SLAMF1, CXCR4, SFT2D1, SH3KBP1, SPOCK2, CDKN1B, GNPTAB, and SNRPA1—demonstrated superior independent prognostic value, with area under the curve (AUC) values of 0.707, 0.698, and 0.724 for 1-, 3-, and 5-year survival predictions, respectively. Immunological profiling revealed that high-risk patients exhibited a prototypical ‘cold’ tumor phenotype, marked by significantly diminished infiltration of stromal and effector immune cells (e.g., CD8+ T cells) and a concomitant enrichment of immunosuppressive populations, including myeloid-derived suppressor cells (MDSCs) and M2-type tumor-associated macrophages (M2-TAMs). Furthermore, high-risk tumors displayed elevated immune dysfunction and exclusion scores alongside reduced IFN-γ and PD-L1 expression, suggesting limited sensitivity to ICI therapy.ConclusionsThis study established a robust prognostic framework based on single-cell resolved T-cell metabolic signatures. Our findings provide a reliable independent prognostic tool for OC and elucidate the intricate interplay between metabolic reprogramming and the immunosuppressive microenvironment. These results offer critical mechanistic insights for prognostic stratification and the development of personalized combinatorial immunotherapies in OC.

Periodontitis and psoriasis are two prevalent conditions that are bidirectionally associated. However, the molecular basis remains poorly understood. This study utilized bioinformatics approaches to investigate the common diagnostic genes and shared mechanisms of periodontitis and psoriasis. Classical datasets for periodontitis and psoriasis were sourced from the GEO database. Differentially expressed genes (DEGs) analysis, weighted gene co-expression network analysis (WGCNA), protein-protein interaction (PPI) network analysis, and two machine learning algorithms were used to screen common biomarkers. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses were utilized to explore biological functions. CIBERSO\ RT was used to assess the immune microenvironment. Transcription factor (TF)-gene and gene-miRNA regulatory networks were analyzed using NetworkAnalyst. 24 DEGs and 333 disease-related genes were identified. Next, three biomarkers (CXCR4, SASH3, and LYN) were identified among the 12 genes shared between DEGs and WGCNA using machine learning. RT-qPCR analysis confirmed the elevated expression of the three shared genes in both conditions. We further constructed a nomogram model and validated it using ROC curves. Immune infiltration analysis revealed a significant association between the three common biomarkers and cellular immune dysregulation. CXCR4, SASH3, and LYN are common biomarkers for psoriasis and periodontitis. Additionally, we proposed immune patterns, TF-gene, and gene-miRNA regulatory networks between the two diseases, which could provide new insights for future studies. Key Points • CXCR4, SASH3, and LYN were identified as shared diagnostic biomarkers for periodontitis and psoriasis, validated through bioinformatics and machine learning approaches. • A robust diagnostic model using the three biomarkers demonstrated high accuracy. • Regulatory networks involving key TFs and miRNAs suggest similar mechanisms between periodontitis and psoriasis.

Exploring potential biomarkers of NETosis-Related genes in spinal cord injury through machine learning and multi-omics analysis.

Dysregulation of immune checkpoint proteins in hepatocellular carcinoma: Impact on metabolic reprogramming

Gliomas are the most aggressive primary malignancies of the central nervous system (CNS) and exhibit marked heterogeneity that is closely associated with metabolic reprogramming. Emerging evidence underscores the pivotal role of lactylation modifications in shaping the tumor microenvironment (TME) and facilitating glioma progression. This study aimed to systematically identify key lactylation-related genes (LRGs), elucidate their functional roles and associated pathways, and explore their potential as novel therapeutic targets using multi-omics data. We combined various datasets from the TCGA, GEO, and CGGA databases, including RNA-seq, single-cell RNA sequencing (scRNA-seq), and spatial transcriptomics. Key LRGs were identified through a multi-step analytical pipeline that involved processing scRNA-seq data using (Seurat, scoring), cell-type-specific lactylation scoring (AUCell), high-dimensional weighted gene co-expression network analysis (hdWGCNA) and applying rigorous machine learning-based feature selection utilizing 10 algorithms and 101 combinatorial strategies. We comprehensively assessed the prognostic value associated with the immune microenvironment, and spatiotemporal heterogeneity of the prioritized RAN. Functional validation was executed using shRNA-mediated knockdown in glioma cell lines, including LN229, U87, and U251, while evaluating proliferation (CCK-8, colony formation, EdU), migration (wound healing), invasion (Transwell), and pathway activity (using western blot). scRNA-seq analysis revealed distinct lactylation enrichment patterns across glioma cell types, with malignant cells exhibiting the highest scores. hdWGCNA identified a gene module (royal blue) strongly correlated with lactylation activity (correlation = 0.75). The intersection of this module with a curated set of LRGs yielded 22 candidate genes. Subsequent machine learning analysis using (ENet, α = 0.4) prioritized six core LRGs (PDAP1, ALYREF, CBX3, MAGOH, RAN, TMSB4X). RAN, an understudied gene in glioma, was selected for further investigation. High RAN expression correlated significantly with poor patient prognosis, reduced immune cell infiltration (assessed by ESTIMATE, CIBERSORT, xCell, ssGSEA), and distinct spatiotemporal heterogeneity within tumors (analyzed using spatial transcriptomics, Monocle2). Glioma cell invasion, migration, colony formation, and proliferation were all markedly inhibited by RAN knockdown. Mechanistically, reduced p-AKT levels following knockdown and functional rescue with a PI3K/AKT activator (SC79) indicate that RAN increased these malignant traits by activating the PI3K/AKT signaling pathway. Our study established lactylation modifications as a crucial regulator of the TME and glioma progression. Through integrative multi-omics analysis and robust machine learning techniques, we determined that RAN was a novel lactylation-associated gene. RAN is a potent, independent prognostic biomarker that promotes glioma malignancy via the PI3K/AKT pathway. Our results demonstrate RAN as a prospective therapeutic target and establish a novel framework for individualized therapy for glioma.