Diabetic nephropathy (DN) is among the most frequent complications of diabetes and the first cause of end-stage renal disease. Despite being successful in animal models, the majority of clinical trials for novel drugs targeting DN failed. This lack of translational value may in part be due to an inadequate comparability of human disease and animal models that often capture only a few aspects of disease. Here we overcome this limitation by developing a multimolecular non-invasive humanized readout of DN based on urinary peptidomics. The diseasemodified urinary peptides of two type 2 diabetic (T2D) DN mouse models were identified and compared with previously validated urinary peptide markers of DN in humans to generate a classifier composed of 21 ortholog peptides. This classifier predicted the response to disease and treatment with inhibitors of the renin-angiotensin system (RASi) in mice. The humanized classifier was significantly correlated with glomerular lesions. Using a human T2D validation cohort consisting of 207 patients, the classifier also distinguished between patients with and without DN, and response to RASi. Our approach demonstrates that a combination of multiple molecular features similar in both human and animal disease could provide a step change in translational drug discovery research in T2D-DN nephropathy.