In vivo CEST MRI data can include contributions from a vast array of metabolites, mobile proteins and peptides, and immobile macromolecules amongst others. Detecting which components are present in any given dataset is a major challenge. Here, as a first start to address the problem, we have used a machine learning approach to classify a CEST dataset acquired from brain metabolite phantoms. The classifier was successful in all cases and was shown to be robust to a moderate level of noise. The results demonstrate this is a promising technique that could potentially quantify molecular contributions in vivo.