Formalin-fixed paraffin-embedded (FFPE) tissue has been the gold standard for routine pathology for general and cancer postoperative diagnostics. Despite robust histopathology, immunohistochemistry, and molecular methods, accurate diagnosis remains difficult for certain cases. Overall, the entire process can be time consuming, labor intensive, and does not reach over 90% diagnostic sensitivity and specificity. There is a growing need in onco-pathology for adjunct novel rapid, accurate, reliable, diagnostically sensitive, and specific methods for high-throughput biomolecular identification. Lipids have long been considered only as building blocks of cell membranes or signaling molecules, but have recently been introduced as central players in cancer. Due to sample processing, which limits their detection, lipid analysis directly from unprocessed FFPE tissues has never been reported. METHODS: We present a proof-of-concept with direct analysis of tissue-lipidomic signatures from FFPE tissues without dewaxing and minimal sample preparation using water-assisted laser desorption ionization mass spectrometry and deep-learning. RESULTS: On a cohort of difficult canine and human sarcoma cases, classification for canine sarcoma subtyping was possible with 99.1% accuracy using "5-fold" and 98.5% using "leave-one-patient out," and 91.2% accuracy for human sarcoma using 5-fold and 73.8% using leave-one-patient out. The developed classification model enabled stratification of blind samples in <5 min and showed >95% probability for discriminating 2 human sarcoma blind samples. CONCLUSION: It is possible to create a rapid diagnostic platform to screen clinical FFPE tissues with minimal sample preparation for molecular pathology.