OBJECTIVE: Lipidomic changes were causally linked to metabolic diseases, but the scenario for colorectal cancer (CRC) is less clear. We investigated the CRC lipidome for putative tumour-specific alterations through analysis of three independent retrospective patient cohorts from two clinical centers, to derive a clinically useful signature. DESIGN: Quantitative comprehensive lipidomic analysis was performed by direct infusion electrospray ionization coupled to tandem mass spectrometry (ESI-MS/MS) and high-resolution mass spectrometry (HR-MS) on matched non-diseased mucosa and tumor tissue in a discovery cohort (n=106). Results were validated in two independent cohorts (n=28, and n=20), associated with genomic and clinical data, and lipidomic data from a genetic mouse tumor model (Apc1638N). RESULTS: Significant differences were found between tumor and normal tissue for glycero-, glycerophospho- and sphingolipids in the discovery cohort. Comparison to the validation collectives unveiled that glycerophospholipids showed high interpatient variation and were strongly affected by preanalytical conditions, whereas glycero- and sphingolipids appeared more robust. Signatures of sphingomyelin (SM) and triacylglycerol (TG) species significantly differentiated cancerous from non-diseased tissue in both validation studies. Moreover, lipogenic enzymes were significantly upregulated in CRC, and FASN gene expression was prognostically detrimental. The TG profile was significantly associated with post-operative disease-free survival and lymphovascular invasion, and was essentially conserved in murine digestive cancer, but not associated with microsatellite status, KRAS or BRAF mutations, or T-cell infiltration. CONCLUSION: Analysis of the CRC lipidome revealed a robust TG-species signature with prognostic potential. A better understanding of the cancer-associated glycerolipid and sphingolipid metabolism may lead to novel therapeutic strategies.