Abstract: SA-PO425

Molecular Re-Classification of CKD Based on Kidney Transcriptomics Profiles

Session Information

Category: Chronic Kidney Disease (Non-Dialysis)

  • 302 CKD: Estimating Equations, Incidence, Prevalence, Special Populations

Authors

  • Reznichenko, Anna, AstraZeneca, Gothenburg, Sweden
  • Liles, John T., Gilead Sciences, Inc., Foster City, California, United States
  • Patel, Uptal D., Gilead Sciences, Inc., Foster City, California, United States
  • Breyer, Matthew D., Eli Lilly and Company, Indianapolis, Indiana, United States
  • Duffin, Kevin L., Eli Lilly and Company, Indianapolis, Indiana, United States
  • Moreno Quinn, Carol Patricia, MedImmune, Cambridge, United Kingdom
  • Magnone, Maria chiara, AstraZeneca, Gothenburg, Sweden
  • Kretzler, Matthias, University of Michigan, Ann Arbor, Michigan, United States
  • Nair, Viji, University of Michigan, Ann Arbor, Michigan, United States
  • Eddy, Sean, University of Michigan, Ann Arbor, Michigan, United States
  • Wei, Tao, Eli Lilly and Company, Indianapolis, Indiana, United States
  • Slidel, Tim, MedImmune, Cambridge, United Kingdom
  • Ju, Wenjun, University of Michigan, Ann Arbor, Michigan, United States
  • Conway, James, MedImmune, Gaithersburg, Maryland, United States
  • Badal, Shawn S., Gilead Sciences, Inc., Foster City, California, United States
  • Wesley, Johnna D, Novo Nordisk Research Center, Seattle, Seattle, Washington, United States

Group or Team Name

  • Renal Precompetitive Consortium (RPC2)
Background

The CKD population is highly heterogeneous and includes a wide range of etiologies with a multitude of underlying molecular processes in the kidney. Current clinical classification of CKD into five stages based on GFR and albuminuria is agnostic to the disease heterogeneity and intrarenal biology, and thus is antithetical to the personalized medicine (PM) concept. We re-classified a CKD patient population based on the kidney molecular profiles, consistent with PM initiatives.

Methods

From clinically indicated renal biopsies in 165 ERCB cohort participants, transcriptomics profiles were generated using Affymetrix U133 platforms. Self-Organizing Maps (SOM), an unsupervised neural network machine learning algorithm, was used to stratify CKD population by clustering cases with similar transcriptomics profiles. Gene Ontology, pathway, and Gene Set Enrichment analyses were performed to identify key molecular mechanisms per SOM cluster.

Results

Using SOM, we identified four distinct patient clusters within the topological map of renal transcriptomics data structure. Relating these molecular clusters back to the current classification revealed the lack of overlap with CKD stages, thus demonstrating that SOM clusters represent a novel characterization beyond clinical classification. The SOM clusters were also not explained by CKD etiology, confirming the hypothesis of disease heterogeneity at the molecular level. Enrichment analyses showed that the SOM clusters differed in terms of biological pathways in the kidney including inflammation, metabolism, cell signaling and apoptosis.

Conclusion

Molecular re-classification may help realize the potential of PM for CKD. Elucidation of the molecular drivers of population clustering can lead to new biological hypotheses, therapeutic targets, and cluster-specific biomarkers that would enable PM-based regimens.

CKD molecular re-classification