Raman spectroscopy can provide a molecular-level signature of the biochemical composition and structure of cells with excellent spatial resolution and could be useful to monitor changes in composition for early stage and non-invasive cancer diagnosis, both ex-vivo and in vivo. In particular, the fingerprint spectral region (400–1,800 cm-1) has been shown to be very promising for optical biopsy purposes. However, limitations to discrimination of dysplastic and inflammatory processes based on the fingerprint region still persist. In addition, the Raman spectral signal of dysplastic cells is one important source of misdiagnosis of normal versus pathological tissues. The high wavenumber region (2,800–3,600 cm-1) provides more specific information based on N-H, O-H and C-H vibrations and can be used to identify the subtle changes which could be important for discrimination of samples. In this study, we demonstrate the potential of the highwavenumber spectral region by collecting Raman spectra of nucleoli, nucleus and cytoplasm from oral epithelial cancer (SCC-4) and dysplastic (DOK) cell lines and from normal oral epithelial primary cells, in vitro, which were then analyzed by area under the curve as a method to discriminate the spectra. In this region, we will show the discriminatory potential of the CH vibrational modes of nucleic acids, proteins and lipids. This technique demonstrated more efficient discrimination than the fingerprint region when we compared the cell cultures.