of endometrial cancer lesions exhibiting high (n=12) or low (n=14) tumour blood flow at DCE-MRI. Univariate and multivariate statistics (ortho-PLS-DA, a random forest (RF) classifier and hierarchical clustering) and receiver operating characteristic (ROC) curves were used to establish a panel for potentially discriminating tumours with high versus low blood flow. Results: Tumour blood flow is associated with specific metabolomic signatures. Ortho-PLS-DA and RF classifier resulted in well-defined clusters with an out-of-bag error lower than 8%. We found 28 statistically significant molecules (False Discovery Rate corrected p<0.05). Based on exact mass, retention time and isotopic distribution we identified 9 molecules including resolvin D and specific lysophospholipids associated with blood flow, and hence with a potentially regulatory role relevant in endometrial cancer. Conclusions: Tumour flow parameters at DCE-MRI quantifying vascular tumour characteristics are reflected in corresponding metabolomics signatures and highlight disease mechanisms that may be targetable by novel therapies.