Most of the techniques used for medical diagnosis, apply methods of analysis to identify certain substances called biomarkers. These techniques generally have the disadvantages of being laborious, invasive, and dependent on the physician’s experience. Raman spectroscopy is projected as a technique capable of identifying biomarkers in a noninvasive, simple and economical way. The analysis of the spectroscopic results by means of multivariable mathematical techniques would allow to eliminate the subjective interpretation of the results and therefore contribute to objective and more reliable diagnoses. The tumor suppressor wild type p53 protein is considered a cancer biomarker. Present in the human body is activated when cellular damage is detected. The p53 protein acts to protect DNA integrity: repairing the damage or inducing cellular death. When p53 do not respond correctly, the damage is not arrested, and tumor growth is developed. Mutations in p53 are related to inactivation of the wild type and therefore the presence of tumors. In this work, Raman spectra of wild type and mutants p53 were obtained through a micro-spectrometer. The spectra were analyzed by multivariate methods. Principal component analysis and support vector machine algorithms showed that it is possible to discriminate between the wild and mutant type of this biomarker with an accuracy of 94%. Raman spectra of wild type p53 at different concentrations were used to estimate the limit of the detection of this protein by means of partial least squares regression. The limit of detection was found as low as 0.946 μM without additional reagents.