Early and accurate detection of acute kidney injury (AKI) is needed to prevent the progression to chronic kidney disease and to improve outcome. Here we used capillary electrophoresis-mass spectrometry to identify urinary peptides predictive of AKI in a training set of 87 urine samples longitudinally collected from patients in an intensive care unit. Within this patient cohort, 16 developed AKI while 14 maintained normal renal function. The sequence of twenty peptides significantly associated with AKI was identified. They were found to be degradation products of six proteins. These formed a diagnostic pattern. Peptides of albumin, α-1-antitrypsin, and β-2-microglobulin were upregulated but fragments of fibrinogen α and collagens 1 α(I) and 1 α(III) were downregulated in AKI. After cross-validation of the training set, a good diagnostic performance of the marker pattern was found with an area under the ROC curve of 0.91. This was confirmed in a blinded validation set of 20 patients in the intensive care unit and 31 allogeneic hematopoietic stem cell transplantation patients, of which 13 had and 18 had not experienced an episode of AKI. In comparison to more established markers of AKI such as serum cystatin C and urinary kidney injury molecule-1, interleukin-18, and neutrophil gelatinase associated-lipocalin, the proteomic marker pattern was found to be of superior prognostic value, detecting AKI up to 5 days in advance of the rise in serum creatinine.