Developing serum proteomics based prediction models of disease progression in ADPKD

Abstract Background: Autosomal Dominant Polycystic Kidney Disease (ADPKD) is the most common genetic cause of kidney failure. Outcome prediction is essential to guide therapeutic decisions. However, currently available models are of limited accuracy. We aimed to examine the potential of serum proteomics for improved risk stratification in ADPKD. Methods: We analyzed the serum proteome of the Screening Cohort (257 patients) using a semi-automated mass spectrometry pipeline and developed models predicting eGFR slope. These models were validated on the Internal/Temporal Cohort (466 patients) and the External Cohort (221 patients). Model performance was assessed by comparing predicted with observed eGFR slopes and compared to imaging- and clinical data based models (e.g. Mayo Imaging Classification (MIC)). Functional implications were explored using gene ontology and pathway analyses. Findings: 398 proteins were identified in the Screening Cohort and a subset of 29 proteins was significantly associated with eGFR slope. Using LASSO-based selection resulted in an optimal protein-based linear prediction model containing six proteins (adjusted R² 0·31). This protein-based model outperformed the Clinical and MIC Models. Combining both molecular and clinical variables further increased the explained variance (adjusted R² 0·34). Predictive value was maintained in the Internal/Temporal Cohort. Also, the models - even though with reduced accuracy - still showed predictive capacity when using EDTA-plasma instead of serum in the External Cohort. Functional enrichment performed on the eGFR slope associated proteins revealed overrepresentation of GO:BP terms related to immune response, lipoprotein levels, metabolic processes and transport. Interpretation: Proteomics is a powerful tool to improve outcome prediction in ADPKD. Importantly, the analyses showed clear added value when combined with currently existing models. Besides, such data harbor valuable information on biological processes associated with disease progression. It will now be important to move towards targeted validation in a prospective study. Funding: Ministry of Science North Rhine-Westphalia, German Research Foundation