Machine learning approach in canine mammary tumour classification using rapid evaporative ionization mass spectrometry.
Journal:
Analytical and bioanalytical chemistry
PMID:
39562368
Abstract
Rapid evaporative ionization mass spectrometry (REIMS) coupled with a monopolar handpiece used for surgical resection and combined with chemometrics has been previously explored by our research group (Mangraviti et al. in Int J Mol Sci 23(18):10562, 2022) to identify several mammary gland pathologies. Here, the increased sample size allowed the construction of three statistical models to distinguish between benign and malignant canine mammary tumours (CMTs), facilitating a more in-depth investigation of changes in cellular metabolic phenotype during neoplastic transformation and biological behaviour. The results demonstrate that REIMS is effective in identifying neoplastic tissues with an accuracy of 97%, with differences in MS spectra characterized by the relative abundance of phospholipids compared to triglycerides more commonly identified in normal mammary glands. The increased rate of phospholipid synthesis represents an informative feature for tumour recognition, with phosphatidylcholine and phosphatidylethanolamine, the two major phospholipid species identified here together with sphingolipids, playing a crucial role in carcinogenesis. REIMS technology allowed the classification of different histotypes of benign CMTs with an accuracy score of 95%, distinguishing them from normal glands based on the increase in sphingolipids, glycolipids, phospholipids, and arachidonic acid, demonstrating the close association between cancer and inflammation. Finally, dysregulation of fatty acid metabolism with increased signalling for saturated, mono- and polyunsaturated fatty acids characterized the metabolic phenotype of neoplastic cells and their malignant transformation, supporting the increased formation of new organelles for cell division. Further investigations on a more significant number of tumour histotypes will allow for the creation of a more extensive database and lay the basis for how understanding metabolic alterations in the tumour microenvironment can improve surgical precision.