Work

Identification of complement system-related biomarkers in immunoglobulin A nephropathy through machine learning and single-cell data analysis

Year: 2026

Type: article

Abstract: Immunoglobulin A nephropathy (IgAN) is associated with complement dysregulation, but complement-related molecular markers remain incompletely defined. We integrated bulk transcriptomic datasets (GSE93798 training; GSE37460 validation), a single-cell RNA-seq dataset (GSE171314), machine-learning feature selection, functional enrichment, immune deconvolution, and a preliminary qRT-PCR assessment to identify candidate complement-system-related biomarkers. Among 25 differentially expressed complement-related genes, LASSO, SVM-RFE, and Boruta jointly prioritized A2M, C3AR1, COLEC11, SERPINA1, and IL1B; A2M, C3AR1, and SERPINA1 showed consistent cross-cohort expression patterns and AUCs > 0.7. A three-gene ANN classifier achieved an apparent training AUC of 1.000 and an external-validation AUC of 0.959, although overfitting remains possible. Functional analyses associated the candidates with ECM-receptor interaction and propionate-metabolism signatures. CIBERSORT and exploratory single-cell analyses suggested macrophage-related associations, but the single-cell dataset included only one healthy control. Archived summary-level qRT-PCR measurements from 5 patients with IgAN and 5 living-donor controls showed higher mean A2M and C3AR1 expression in IgAN tissue; these measurements are presented descriptively because participant-level raw values and the original statistical-test output were not available for independent reanalysis. These findings support A2M, C3AR1, and SERPINA1 as candidate IgAN biomarkers requiring validation in larger, clinically annotated cohorts.

Source: Scientific Reports

DOI: 10.1038/s41598-026-77340-9