Abstract: SA-PO245
Modeling Progression of Uremic Vasculopathy Using Machine Learning
Session Information
- CKD-MBD: Basic and Translational
October 26, 2024 | Location: Exhibit Hall, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Bone and Mineral Metabolism
- 501 Bone and Mineral Metabolism: Basic
Authors
- Gaweda, Adam E., University of Louisville, Louisville, Kentucky, United States
- Lederer, Eleanor D., The University of Texas Southwestern Medical Center, Dallas, Texas, United States
- Brier, Michael E., University of Louisville, Louisville, Kentucky, United States
Background
KDIGO guidelines for Chronic Kidney Disease Mineral Bone Disorder (CKD-MBD) target Phosphate, Parathyroid Hormone, and Calcium as the primary clinical outcomes. To prevent vascular end organ damage due to CKD-MBD, we need better understanding of the mechanism behind uremic vasculopathy. We perform nonlinear analysis of biochemical parameters in the Chronic Renal Insufficiency Cohort (CRIC) and their association with coronary artery calcification (CAC).
Methods
We abstracted 93 biochemical parameters from the CRIC data set and performed 10 data fitting sessions using Multilayer Perceptron Neural Network model, splitting the data into training (70%) and testing (30%) sets. Using sensitivity analysis, we selected the most important parameters from the trained model. We ranked these parameters and analyzed their association with CAC at different CKD stages. Data analysis was performed in IBM SPSS.
Results
The average CAC prediction RMSE was 0.08 (training) and 0.11 (testing). At early CKD stages, CAC appears to be influenced by oxidative stress, mitochondrial dysfunction, and alterations in protein translation and post-translational modification (YKL40, ADMA, succinic acid, NAG+, Tiglylglycine). High CAC scores in early CKD correlate with high FGF23, whereas high Calcitriol appears protective. At late CKD stages, high CAC scores reflect impaired tubular function including xanthosine, BTP, and B2M.
Conclusion
Uremic vasculopathy involves various pathways active at different CKD stages. Biomarker discovery enables identifying these pathways to better understand the disease process, leading to better diagnosis, prevention, and treatment. Acknowledgment: CRIC data were provided by NIDDK Central Repository, a program of the National Institute of Diabetes and Digestive and Kidney Diseases.
Regression plot for predicted vs actual CAC score.
Funding
- Veterans Affairs Support