Aberrations in DNA methylation patterns are associated with various diseases, including cancer, and their study may provide valuable signs for early detection of this disease. The aim of this paper is to propose a methodology for the analysis of such epigenetic profiles, with the objective of distinguishing between different lymphoma subtypes. In our analysis, we start from DNA methylation data, we identify differentially methylated regions using a proposed naïve approach and an established statistical approach (Dispersion Shrinkage for Sequencing algorithm [13]), perform feature selection using variance thresholding and the Minimum Redundancy Maximum Relevance method [7] and apply non-linear feature extraction with autoencoders. Lastly, we train three classifiers (XGBoost, Support Vector Machines, and Artificial Neural Networks) on the obtained, reduced feature sets. The results show that a particular combination of DMR extraction procedure, feature selection/extraction and classifier performs best on both an internal test set (held-out from the original data) and an external test set. However, considering the limited size of the input data (only 30 samples), these findings should be considered preliminary and require further validation in future studies.