Using the CKD273 urinary peptide classifier, researchers simulated treatment effects in silico for 935 CKD patients, finding personalized drug combinations that significantly reduced disease progression scores — paving the way for peptide-guided precision medicine.
CKD273 score: 0.57 → 0.039Personalized in silico drug combination selection based on urinary peptide profiles dramatically reduced kidney disease progression scores across 935 patients
What the researchers found
In 935 CKD patients with validated kidney event outcomes:
- CKD273 peptide classifier confirmed as predictor of major adverse kidney events (≥40% eGFR decline or kidney failure)
- In silico simulation of 6 interventions: mineralocorticoid receptor antagonist, SGLT2 inhibitor, GLP-1 receptor agonist, ARB, olive oil diet, and exercise
- Simulated optimal interventions reduced median CKD273 from 0.57 to 0.039 (P < 0.0001)
- The combination of all available treatments was NOT the most frequently predicted optimal intervention — personalization mattered
- Patients with higher baseline scores required more complex combinations
- Approach uses individual urinary peptide profiles + known treatment effects on peptide abundance
Why it matters
This study demonstrates that urinary peptide patterns can guide personalized treatment selection — a major advance toward precision medicine in kidney disease. Rather than giving all patients the same drugs, the peptide profile identifies which combination would work best for each individual. If validated prospectively, this could transform how kidney disease is treated.
How the study worked
Retrospective cohort of 935 CKD patients with urinary peptidomic data. CKD273 classifier scores were validated against major adverse kidney events over median 1.5-year follow-up. In silico treatment simulation used: (1) individual baseline urinary peptide profiles, and (2) previously defined fold changes in peptide abundance from clinical trials of four drugs, dietary intervention, and exercise. Simulated CKD273 scores were recalibrated against trial outcomes. Optimal single and combination treatments were predicted for each patient.
What this study cannot tell us
The treatment simulations are based on average drug effects from prior trials, which may not perfectly predict individual responses. The study is retrospective and computational — the in silico predictions need prospective clinical validation. The model assumes treatment effects on peptides are additive, which may not always be true. A prospective validation trial is planned but not yet completed.
How to read the evidence
This is a computational study using a retrospectively validated peptide classifier. While the CKD273 classifier itself is clinically validated, the treatment prediction model is novel and requires prospective validation in a clinical trial.
When this study was published
Published in 2025, this is at the cutting edge of peptide biomarker-guided precision medicine. A prospective validation trial is being planned.
The bigger picture
Urinary peptidomics represents a frontier in precision medicine, where patterns of hundreds of peptide fragments in urine serve as molecular fingerprints of disease. The CKD273 classifier was previously validated in a prospective clinical trial and is commercially available. This study extends its use from diagnosis to treatment guidance — a paradigm shift from 'one-size-fits-all' to peptide-guided personalized therapy.
Questions still open
- Will the predicted optimal interventions actually improve kidney outcomes when tested in a prospective trial?
- Can this peptide-guided approach be extended to other diseases where urinary peptide profiles change with treatment?
- How often should urinary peptide profiles be re-measured to adjust treatment strategies over time?
Common questions
What is the CKD273 urinary peptide classifier?
How does this approach personalize kidney disease treatment?
Read the original research
In silico prediction of optimal multifactorial intervention in chronic kidney disease.
Journal of translational medicine, 23(1), 943
Citation
Latosinska, Agnieszka; Mina, Ioanna K; Nguyen, Thi Minh Nghia; Golovko, Igor; Keller, Felix; Mayer, Gert; Rossing, Peter; Staessen, Jan A; Delles, Christian; Beige, Joachim; Glorieux, Griet; Clark, Andrew L; Schanstra, Joost P; Vlahou, Antonia; Peter, Karlheinz; Rychlík, Ivan; Ortiz, Alberto; Campbell, Archie; Rupprecht, Harald; Persson, Frederik; Mischak, Harald; Siwy, Justyna. (2025). In silico prediction of optimal multifactorial intervention in chronic kidney disease.. Journal of translational medicine, 23(1), 943. https://doi.org/10.1186/s12967-025-06977-3