Abstract: FR-PO521
NMR Metabolomic Profiling in Distinguishing Kidney Stone Formers from Controls
Session Information
- Bone and Mineral Metabolism: Basic
October 26, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Bone and Mineral Metabolism
- 401 Bone and Mineral Metabolism: Basic
Authors
- Denic, Aleksandar, Mayo Clinic, Rochester, Minnesota, United States
- Vuckovic, Ivan, Mayo Clinic, Rochester, Minnesota, United States
- Vaughan, Lisa E., Mayo Clinic, Rochester, Minnesota, United States
- Enders, Felicity T., Mayo Clinic, Rochester, Minnesota, United States
- Macura, Slobodan, Mayo Clinic, Rochester, Minnesota, United States
- Lieske, John C., Mayo Clinic, Rochester, Minnesota, United States
- Rule, Andrew D., Mayo Clinic, Rochester, Minnesota, United States
Background
NMR-based urine metabolomics has shown considerable potential in biomarker discovery. It is not fully understood why some individuals develop kidney stones and others do not. We hypothesized that certain urine metabolites can distinguish kidney stone formers from controls.
Methods
We used high resolution NMR spectroscopy to study urines of 255 adult first-time incident symptomatic stone formers (SF), and 255 age and sex-matched controls. This was split into a training dataset (2/3, n=171 SF/control each) and a test dataset (1/3, n=84 SF/control each). NMR spectra were normalized by total spectrum area. Sex and metabolites from NMR data were used in PLS discrimination analysis to fit a model that distinguishes stone formers from controls in the training dataset. Variables with VIP>0.8 were selected for inclusion and cross-validation was used to determine number of components in the final model.
Results
Among the 108 metabolites included in the final model, those with the strongest positive association with SF were 1,6-Anhydro-D-glucose, glucuronate, lactose and an unknown compound, while those with the strongest negative association were 2 unknown compounds and scyllo-inositol. The area under the curve for detecting stone formers with the metabolite panel was 0.66 in the training dataset and 0.57 in the validation dataset (Figure). Sensitivity and specificity were 78%/50% and 76%/36% in the training and validation sets, respectively.
Conclusion
This study found that a panel of urine metabolites could potentially discriminate those at risk for symptomatic kidney stones. Further studies are needed to identify the unknown compounds. Understanding the differential expression of these metabolites could shed new insights into the biology of kidney stone formation.
Funding
- NIDDK Support