Abstract: FR-OR12
Identifying Diabetic Kidney Disease Signatures in the Nuclei of the Tubular Epithelium Using a Novel Deep Learning Approach
Session Information
- Diabetic Kidney Disease: Back to the Basics
November 05, 2021 | Location: Simulive, Virtual Only
Abstract Time: 04:30 PM - 06:00 PM
Category: Diabetic Kidney Disease
- 601 Diabetic Kidney Disease: Basic
Authors
- Winfree, Seth, Indiana University School of Medicine, Indianapolis, Indiana, United States
- Barwinska, Daria, Indiana University School of Medicine, Indianapolis, Indiana, United States
- Talukder, Niloy, Indiana University School of Medicine, Indianapolis, Indiana, United States
- Eadon, Michael T., Indiana University School of Medicine, Indianapolis, Indiana, United States
- Dagher, Pierre C., Indiana University School of Medicine, Indianapolis, Indiana, United States
- Hasan, Mohammad Al, Indiana University School of Medicine, Indianapolis, Indiana, United States
- El-Achkar, Tarek M., Indiana University School of Medicine, Indianapolis, Indiana, United States
Group or Team Name
- KPMP
Background
Diabetic nephropathy (DN), a leading cause of end stage kidney disease (ESKD) is generally viewed as a glomerular disease. However, progression of DN towards ESKD correlates best with tubular pathology and fibrosis. Due to the spatial complexity of the human kidney, which includes many cell types, it is a challenge to capture the biology at the single cell level. While there is a growing body of information on the molecular phenotype of DN at the single cell level using omics approaches on dissociated tissue, there is little information on cellular changes in intact kidney tissue.
Methods
We used a 3D nuclei image-based deep learning approach to uncover spatially resolved single cell signatures of DN. 3D Imaging datasets were collected from fluorescently labeled human reference nephrectomy samples and biopsies from patients with DN. Using Volumetric Tissue Exploration and Analysis (VTEA) and cell-type markers, a 3D nuclei image dataset was generated from reference nephrectomies and used to train a custom Convolutional Neural Network (CNN). A second 3D nuclei image dataset was generated from images of biopsies taken from patients with DN and classified with the trained CNN.
Results
We generated a 3D nuclei image library from DN tissue secured from the NIDDK/Kidney Precision Medicine Project. We used our nuclei-based CNN classification of renal cells to uncover unique classes of renal epithelium and identify novel single cell image-based signature in DN. Using VTEA, we were able to spatially localize these novel classes of renal epithelium and assess correlation with injury and renal structures for a spatially resolved 3D nuclei image-based signature of DN.
Conclusion
Our work demonstrates that 3D nuclei images from renal cells allows for the identification of DN signatures. These data further suggest that in addition to glomeruli, the tubular epithelium plays a role in DN. Our work underlines the potential of using machine learning and deep learning approaches to automatically uncover new cell types which may emerge due to changes occurring in diabetes, while maintaining their spatial context. Thus, our work can provide insight into the cellular changes in intact kidney tissue during progression in DN.
Funding
- NIDDK Support