Classifying Raman Spectra of Colon Cells by Principal Component Analysis—Linear Discriminant Analysis and Partial Least Squares—Linear Discriminant Analysis Methods

Colorectal cancer is one of the most commonly diagnosed cancers in developed countries. Although the gold-standard diagnosis technique is the histological analysis of colon biopsies, it is important to investigate different diagnostic tools because the microscope examination of stained tissues provides indications partially depending on the experience of the pathologist. This study reports a Raman-spectroscopy-based analysis of healthy and cancerous colon cells to detect biochemical differences at the subcellular level and discriminate the former from the latter. FHC and CaCo-2 cell lines were used to model healthy and cancerous cells, respectively. The comparison of the Raman spectra measured inside subcellular volumes including the nucleus (nucleus spectra) and excluding it (cytoplasm spectra), as well as principal component analysis and partial least squares analysis of these spectra, suggest that the differences between the spectra of healthy and cancerous cells are very small, and they mainly involve the different relative content of lipids and nucleic acid components. The relative intensity of lipid peaks is higher in the Raman spectra of healthy samples, while nucleic acid peaks show higher relative intensity in the spectra of cancer cells. Linear discriminant analysis of a few principal components and partial least squares components was used to estimate the classification accuracy of a set of Raman spectra measured inside nucleus and cytoplasm. Both methods are able to classify unknown cells with excellent accuracy (100% and 96%, respectively). The findings of this study confirm the general applicability of subcellular Raman analysis in clinical practice for diagnosis of cytological samples.