Figure 3.

Prediction of cancer relapse using the NMR metabolomic fingerprint. The model calculated on NOESY1D spectra, discriminating patients free from eCRC disease relapse (light blue, dots, n = 65) and patients with mCRC (red dots, n = 75) using the PCA-CA-kNN analysis, was used to classify patients who developed disease recurrence. Twenty (green triangles) out of 29 patients were correctly classified as “metastatic” disease, the other nine patients (green squares) were mis-classified as early disease. This model provides an overall prediction accuracy of 69%.