Background As a critical hallmark of systemic lupus erythematosus, lupus nephritis (LN) stands as a major organ-threatening condition driven by sophisticated immunological imbalances. Aberrant glycosylation has been implicated in autoimmune pathogenesis, yet its diagnostic and immunological roles in LN remain unclear. Methods The Gene Expression Omnibus repository served as the primary source for the LN-related microarray datasets utilized in this study. Upon correcting for batch effects, we executed integrated analyses to detect differentially expressed glycosylation-associated markers by integrating the limma-based approach with weighted gene coexpression network analysis-driven modeling. Protein–protein interaction networks and functional enrichment were conducted to reveal hub genes. To pinpoint stable diagnostic signatures, we integrated three computational frameworks: least absolute shrinkage and selection operator, support vector machine-recursive feature elimination, and random forest. Immune infiltration was analyzed by single-sample gene set enrichment analysis, CIBERSORT, and MCP-Counter, and regulatory networks were predicted using NetworkAnalyst and miRNet. Molecular subgroups were delineated using the ConsensusClusterPlus package, with their biological implications subsequently explored through Gene Ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses. Results Three diagnostic genes—C1QA, CD14, and CTSS—were identified with high diagnostic accuracy (area under the curve > 0.7) and strong immune correlations. These genes were enriched in immune response–related pathways, such as phagocytosis, Toll-like receptor signaling, and cytokine–cytokine receptor interactions. Two immune-related molecular subtypes of LN were identified, exhibiting distinct immune infiltration patterns and activation states. Conclusion This study established a robust glycosylation-related diagnostic model for LN and uncovered two immunologically distinct molecular subtypes. The identified biomarkers (C1QA, CD14, and CTSS) may serve as potential diagnostic indicators and provide insights into LN immune pathogenesis. <Table Float="No" ID="Taba"> <tgroup cols="2"> <colspec align="left" colname="c1" colnum="1" /> <colspec align="left" colname="c2" colnum="2" /> <tbody> <row> <entry align="left" nameend="c2" namest="c1"> Key Points • A glycosylation-related diagnostic model for lupus nephritis (LN) was established by integrating multicohort transcriptomic data and three machine learning algorithms (LASSO, SVM-RFE, and random forest). • Three robust diagnostic biomarkers (C1QA, CD14, and CTSS) were identified, exhibiting strong diagnostic efficacy and close associations with immune activation, phagocytosis, and inflammatory pathways. • Two immune-related molecular subtypes of LN were defined, characterized by distinct immune infiltration patterns and biological functions, providing new insights into LN heterogeneity and potential therapeutic stratification. </entry> </row> </tbody> </tgroup> </Table>

Construction and validation of a glycosylation-related diagnostic model and immune characterization in lupus nephritis

期刊论文
发表日期：

2026年4月29日

查看全文

Li’nan Zhang,
Yixian Liang

摘要

Background

As a critical hallmark of systemic lupus erythematosus, lupus nephritis (LN) stands as a major organ-threatening condition driven by sophisticated immunological imbalances. Aberrant glycosylation has been implicated in autoimmune pathogenesis, yet its diagnostic and immunological roles in LN remain unclear.

Methods

The Gene Expression Omnibus repository served as the primary source for the LN-related microarray datasets utilized in this study. Upon correcting for batch effects, we executed integrated analyses to detect differentially expressed glycosylation-associated markers by integrating the limma-based approach with weighted gene coexpression network analysis-driven modeling. Protein–protein interaction networks and functional enrichment were conducted to reveal hub genes. To pinpoint stable diagnostic signatures, we integrated three computational frameworks: least absolute shrinkage and selection operator, support vector machine-recursive feature elimination, and random forest. Immune infiltration was analyzed by single-sample gene set enrichment analysis, CIBERSORT, and MCP-Counter, and regulatory networks were predicted using NetworkAnalyst and miRNet. Molecular subgroups were delineated using the ConsensusClusterPlus package, with their biological implications subsequently explored through Gene Ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses.

Results

Three diagnostic genes—C1QA, CD14, and CTSS—were identified with high diagnostic accuracy (area under the curve > 0.7) and strong immune correlations. These genes were enriched in immune response–related pathways, such as phagocytosis, Toll-like receptor signaling, and cytokine–cytokine receptor interactions. Two immune-related molecular subtypes of LN were identified, exhibiting distinct immune infiltration patterns and activation states.

Conclusion

This study established a robust glycosylation-related diagnostic model for LN and uncovered two immunologically distinct molecular subtypes. The identified biomarkers (C1QA, CD14, and CTSS) may serve as potential diagnostic indicators and provide insights into LN immune pathogenesis.

Key Points

• A glycosylation-related diagnostic model for lupus nephritis (LN) was established by integrating multicohort transcriptomic data and three machine learning algorithms (LASSO, SVM-RFE, and random forest).

• Three robust diagnostic biomarkers (C1QA, CD14, and CTSS) were identified, exhibiting strong diagnostic efficacy and close associations with immune activation, phagocytosis, and inflammatory pathways.

• Two immune-related molecular subtypes of LN were defined, characterized by distinct immune infiltration patterns and biological functions, providing new insights into LN heterogeneity and potential therapeutic stratification.