In this work we perform gene expression profiling on tissue specimen obtained from patients with oral squamous cell carcinoma with a twofold aim: i) to identify a limited number of genes that capture perturbations at molecular level dictating the development of a potential disease relapse after remission, and ii) to employ these genes in order to build a classifier that is able to calculate the probability of disease reoccurrence for new patients, subsequently discriminating patients into high and low risk groups based on reoccurrence probability. The proposed analysis yielded 94% overall accuracy, 100% sensitivity and 89% specificity, for discriminating patients with and without a disease relapse.