Systemic proteome profiling for differentiation of primary glomerular diseases

Primary GN refers to a heterogeneous group of kidney diseases marked by immune-mediated injury and reduced renal function. Accurate classification of GN subtypes is essential to inform immunosuppressive treatment decisions and to predict kidney survival. However, reliable diagnostic biomarkers are still lacking.

Advances in high-throughput proteomics now enable the characterisation of thousands of circulating proteins, providing new insights into disease mechanisms and biomarker discovery. While systemic proteome profiling has the potential to distinguish GN subtypes based on their unique protein signatures, previous studies have been hindered by constrained proteomic depth and a reliance on binary comparisons.

To identify distinct protein signatures that discriminate GN subtypes and to assess their potential for non-invasive disease classification, Oh and colleagues performed large-scale proteome profiling of 5416 plasma proteins. The profiling was conducted in a discovery cohort (n=147) and an external validation cohort (n=85) of Korean participants (mean age, 41±13 years; 46% female). 

The study participants included individuals with four GN subtypes – focal segmental glomerulosclerosis (FSGS), IgA nephropathy, minimal change disease and anti-phospholipase A2 receptor (PLA2R) antibody-associated membranous nephropathy – alongside healthy controls. 

Initial proteome profiling identified distinct signatures across biopsy-proven GN subtypes. A machine learning (ML) model was subsequently developed using logistic regression with elastic net regularisation to discriminate GN subtypes based on profiles using 93 proteins. Its performance was evaluated in the independent validation cohort. 

The model was robust in identifying minimal change disease, membranous nephropathy and IgA nephropathy, achieving an area under the receiver operating characteristic curve >0.8. Classification performance for FSGS was limited, identifying fewer than half of the FSGS cases, indicating a lack of decisive systemic signatures for this subtype. The overall model performance remained similar when baseline clinical variables (eGFR and UPCR) were added. 

The findings support the potential of plasma-based proteomics integrated with ML in the differential diagnosis of GN alongside conventional clinical markers. However, there are boundaries to its accuracy and further validation studies are needed. 

The full article can be read here.