Although molecular classification brings interesting insights into breast cancer taxonomy, its implementation in daily clinical care is questionable because of its expense and the information supplied in a single sample allocation is not sufficiently reliable. New approaches, based on a panel of small molecules derived from the global or targeted analysis of metabolic profiles of cells, have found a correlation between activation of de novo lipogenesis and poorer prognosis and shorter disease-free survival for many tumors. We hypothesized that the lipid content of breast cancer cells might be a useful indirect measure of a variety of functions coupled to breast cancer progression. Raman microspectroscopy was used to characterize metabolism of breast cancer cells with different degrees of malignancy. Raman spectra from MDA-MB-435, MDA-MB-468, MDA-MB-231, SKBR3, MCF7 and MCF10A cells were acquired with an InVia Raman microscope (Renishaw) with a backscattered configuration. We used Principal Component Analysis and Partial Least Squares Discriminant Analyses to assess the different profiling of the lipid composition of breast cancer cells. Characteristic bands related to lipid content were found at 3014, 2935, 2890 and 2845 cm(-1), and related to lipid and protein content at 2940 cm(-1). A classificatory model was generated which segregated metastatic cells and non-metastatic cells without basal-like phenotype with a sensitivity of 90% and a specificity of 82.1%. Moreover, expression of SREBP-1c and ABCA1 genes validated the assignation of the lipid phenotype of breast cancer cells. Indeed, changes in fatty acid unsaturation were related with the epithelial-to-mesenchymal transition phenotype. Raman microspectroscopy is a promising technique for characterizing and classifying the malignant phenotype of breast cancer cells on the basis of their lipid profiling. The algorithm for the discrimination of metastatic ability is a first step towards stratifying breast cancer cells using this rapid and reagent-free tool.