Systemic Proteome Profiling to Differentiate Primary Glomerular Diseases.

Area of Science:

• Nephrology
• Proteomics
• Machine Learning

Background:

• Primary glomerulonephritis (GN) is a complex kidney disorder with incomplete pathophysiology understanding.
• Current diagnostic methods have limitations in differentiating GN subtypes using systemic signatures.

Purpose of the Study:

• To identify noninvasive protein signatures for differentiating major primary GN subtypes.
• To gain mechanistic insights into GN pathophysiology through proteomic profiling.
• To evaluate the feasibility of using machine learning models for GN classification.

Main Methods:

• Performed large-scale systemic proteome profiling of 5,416 plasma proteins using Olink Explore HT.
• Utilized discovery (n=147) and validation (n=85) cohorts of Korean participants with four GN subtypes and healthy controls.
• Developed and evaluated a machine learning model (logistic regression with elastic net regularization) for disease classification.

Main Results:

• Plasma proteome profiles distinctly varied among GN subtypes, independent of conventional markers like eGFR.
• The machine learning model achieved an AUROC > 0.8 for differentiating minimal change disease, membranous nephropathy, and IgA nephropathy.
• The model showed high accuracy for minimal change disease (93%) and IgA nephropathy (63%), but limited performance for focal segmental glomerulosclerosis (21%).

Conclusions:

• Distinct systemic proteome signatures were identified for primary GN subtypes.
• Disease subtype significantly influences proteomic variance, complementing clinical markers.
• Machine learning models show promise for proteome-based classification of minimal change disease, membranous nephropathy, and IgA nephropathy.