Mass Spectrometry-based non-targeted lipidomics study for biomarker discovery in clear cell renal cell carcinoma

MALENA MANZI; MARTÍN PALAZZO; MARÍA ELENA KNOTT; NICOLÁS ZABALEGUI; PIERRE BEAUSEROY; PATRICIO YANKILEVICH; MARÍA ISABEL GIMÉNEZ; MARÍA EUGENIA MONGE

Workshop; 4th International Mass Spectrometry School; 2019

Clear cell renal cell carcinoma (ccRCC) is the most common (75%) lethal subtype of RCC, and it is classified into stages I, II, III and IV; the latter being a metastatic cancer. ccRCC is considered a glycolytic and lipogenic tumor (Hsie et al. 2017; Hakimi et al. 2013). More than 30% of cases are incidentally detected by imaging procedures, and 17% of the patients exhibit locally advanced or metastatic RCC at the time of diagnosis when treatment effectiveness (Hu et al. 2012, Graves et al. 2013) is reduced because tumors are inherently resistant to chemotherapy and radiotherapy. However, when the tumor is localized and detected at early stages, the disease is potentially curable by surgery, highlighting the importance of an early diagnosis. Indeed, the discovery of early detection biomarkers is the most promising approach to reduce RCC mortality (Gao et al. 2012). The present work consist in a discovery-based lipid profiling study of serum samples using ultraperformance liquid chromatography coupled to quadrupole-time-of-flight mass spectrometry and machine learning statistical analysis. The cohort included patients with ccRCC stages (stage I, II, III and, IV; n=112) and controls (n=52). For variable selection, machine learning algorithms using support vector machines (SVM) coupled with the least absolute shrinkage and selection operator (Lasso) using balanced sample classes was applied to a 386-feature matrix. Two discriminant lipid-panels for ccRCC detection and early diagnosis were obtained: i) 18 compounds allowed the discrimination of ccRCC patients from controls with 96.8% and 81.4% accuracy in a training set under cross-validation, and in an independent test set, respectively, ii) 26 compounds allowed the discrimination of early stages (I and II) from late stages (III and IV) ccRCC samples with 84.5% accuracy in the training set under cross-validation, and 82.1% accuracy in the classification of stage I ccRCC patients from an independent test set. Interestingly, 15 out of 18 discriminant lipids that belong to the first panel were significant decreased in patients with ccRCC respect to controls (p