Machine Learning and Experimental Validation Ubiquitination-Related Key Genes in Sepsis: The Perspective of Chinese-Western Pharmaceutical Therapeutic Strategies

Abstract

Background and Objectives: Sepsis, a global health challenge, is characterized by a systemic inflammatory response syndrome (SIRS) triggered by infection, resulting in a complex interplay between inflammation and immune suppression. Recent research highlights ubiquitination, a critical post-translational modification, as pivotal in maintaining immune equilibrium and cellular stress responses by modulating protein stability and activity in inflammatory pathways. Despite these insights, significant knowledge gaps persist regarding the specific ubiquitination regulatory network in sepsis, including the spatiotemporal expression of key ubiquitin ligases and deubiquitinases and their link to organ dysfunction progression. These gaps hinder the development of precision therapies targeting the ubiquitination regulatory network. This study, integrating traditional Chinese and Western medicine, aims to elucidate ubiquitination's role in sepsis immune-metabolic reprogramming and perform high-throughput screening of traditional Chinese and small-molecule drugs targeting sepsis-related ubiquitinated proteins. The goal is to identify precise intervention targets and potential treatments, facilitating personalized strategies based on ubiquitination molecular profiling for sepsis patients. Methods: We integrated four independent sepsis cohort transcriptome gene expression datasets (GSE32707, GSE54514, GSE69528, GSE137340) from the GEO database, comprising 423 clinical samples (110 healthy controls and 313 sepsis patients), to establish a framework for analyzing ubiquitination-related genes (UbRGs) across multiple centers. Sepsis Differentially Expressed Genes (DEGs) were identified using the R Limma package with criteria of |logFC| > 0.2 and adj. p < 0.05. These DEGs were intersected with 1,838 UbRGs from the GeneCards database through Venn analysis to pinpoint sepsis-related differentially expressed UbRGs. Functional characteristics were explored using GO/KEGG pathway and gene set enrichment analyses. Key characteristic genes were identified via LASSO regression, Random Forest (RF), and XGBoost machine learning algorithms. CIBERSORT analysis quantified peripheral blood immune cell subsets, and the area under the ROC curve assessed the diagnostic efficacy of candidate genes, leading to the selection of optimal diagnostic markers. Candidate gene expression levels were validated through RT-qPCR in an in vitro sepsis model using LPS-induced THP-1/RAW264.7 macrophage cell lines. UbRGs-related drugs, including traditional Chinese medicine (TCMs) monomers and small molecules, were screened from COREMINE, HERB, and TCMSP databases, with binding capabilities confirmed via molecular docking. Results: This study identified 448 differentially expressed Ubiquitin-Related Genes (UbRGs) associated with sepsis by analyzing multiple sepsis cohorts. (1) Functional annotation revealed significant enrichment of UbRGs in biological processes including protein ubiquitination, NF-κB signaling pathway, and apoptosis. (2) Machine learning analysis identified 11 key UbRGs (e.g., UBE2C, KLHL3, CAPDH) through LASSO regression, Random Forest (RF), and XGBoost, with 7 genes (e.g., E4F1, UBE2C) showing moderate to high correlation with sepsis severity (AUC > 0.7). (3) Immune infiltration analysis indicated a significant increase in mast cells and B cells in sepsis, with a moderate negative correlation between M macrophages and resting CD4+ T cells (r = -0.55, P < 0.05), and activated dendritic cells (r = -0.54, P <0.05). The high-expression sepsis group was enriched in pathways such as NF-κB signaling, apoptosis, and ubiquitin-mediated proteolysis, suggesting a potential role in modulating immune cell functions in sepsis progression. (4) Single-gene functional mechanism analysis highlighted that immune infiltration analysis of 4 characteristic DE-UbRGs (FBXO7, SIAH1, SUMO2, UBE2C) correlated significantly with various immune cells and pathways related to cellular stress response, energy metabolism, and immune activation, emphasizing their importance in inflammatory responses and cell homeostasis. (5) RT-qPCR confirmed decreased mRNA levels of FBXO7, SIAH1, and SUMO2, and increased mRNA level of UBE2C in an in vitro sepsis cell model (THP-1) (P < 0.05). (6) Small-molecule drugs like Sirolimus and Temozolomide exhibited strong binding affinity with the 4 characteristic DE-UbRGs, and TCMs associated with these UbRGs and their active ingredients also demonstrated favorable binding abilities. Conclusion: This study identified key ubiquitination genes associated with sepsis pathology, including inflammation and the immune microenvironment, using machine learning. It also investigated TCMs monomers and small molecule drugs for targeted therapy, laying the groundwork for developing personalized treatment strategies based on ubiquitination molecular typing of sepsis with substantial clinical translation potential.