Abstract
Cervical cancer is still an important issue of public health since it is the fourth most frequent type of cancer in women worldwide. Much effort has been dedicated to combating this cancer, in particular by the early detection of cervical pre-cancerous lesions. For this purpose, this paper reports the use of mass spectrometry coupled with multivariate analysis as an untargeted lipidomic approach to classifying 76 blood plasma samples into negative for intraepithelial lesion or malignancy (NILM, n = 42) and squamous intraepithelial lesion (SIL, n = 34). The crude lipid extract was directly analyzed with mass spectrometry for untargeted lipidomics, followed by multivariate analysis based on the principal component analysis (PCA) and genetic algorithm (GA) with support vector machines (SVM), linear (LDA) and quadratic (QDA) discriminant analysis. PCA-SVM models outperformed LDA and QDA results, achieving sensitivity and specificity values of 80.0% and 83.3%, respectively. Five types of lipids contributing to the distinction between NILM and SIL classes were identified, including prostaglandins, phospholipids, and sphingolipids for the former condition and Tetranor-PGFM and hydroperoxide lipid for the latter. These findings highlight the potentiality of using mass spectrometry associated with chemometrics to discriminate between healthy women and those suffering from cervical pre-cancerous lesions.