Discovery of Novel Digital Biomarkers for Type 2 Diabetic Nephropathy Classification via Integration of Urinary Proteomics and Pathology

BackgroundThe heterogeneous phenotype of diabetic nephropathy (DN) from type 2 diabetes complicates appropriate treatment approaches and outcome prediction. Kidney histology helps diagnose DN and predict its outcomes, and an artificial intelligence (AI)- based approach will maximize clinical utility of histopathological evaluation. Herein, we addressed whether AI-based integration of urine proteomics and image features improves DN classification and its outcome prediction, altogether augmenting and advancing pathology practice. MethodsWe studied whole slide images (WSIs) of periodic acid-Schiff-stained kidney biopsies from 56 DN patients with associated urinary proteomics data. We identified urinary proteins differentially expressed in patients who developed end-stage kidney disease (ESKD) within two years of biopsy. Extending our previously published human-AI-loop pipeline, six renal sub-compartments were computationally segmented from each WSI. Hand-engineered image features for glomeruli and tubules, and urinary protein measurements, were used as inputs to deep-learning frameworks to predict ESKD outcome. Differential expression was correlated with digital image features using the Spearman rank sum coefficient. ResultsA total of 45 urinary proteins were differentially detected in progressors, which was most predictive of ESKD (AUC=0.95), while tubular and glomerular features were less predictive (AUC=0.71 and AUC=0.63, respectively). Accordingly, a correlation map between canonical cell-type proteins, such as epidermal growth factor and secreted phosphoprotein 1, and AI-based image features was obtained, which supports previous pathobiological results. Conclusions: Computational method-based integration of urinary and image biomarkers may improve the pathophysiological understanding of DN progression as well as carry clinical implications in histopathological evaluation. Significance StatementThe complex phenotype of diabetic nephropathy from type 2 diabetes complicates diagnosis and prognosis of patients. Kidney histology may help overcome this difficult situation, particularly if it further suggests molecular profiles. This study describes a method using panoptic segmentation and deep learning to interrogate both urinary proteomics and histomorphometric image features to predict whether patients progress to end-stage kidney disease since biopsy date. A subset of urinary proteomics had the most predictive power in identifying progressors, which could annotate significant tubular and glomerular features related to outcomes. This computational method, which aligns molecular profiles and histology, may improve our understanding of pathophysiological progression of diabetic nephropathy as well as carry clinical implications in histopathological evaluation.