An Acute Coronary Syndrome (ACS) is a set of clinical signs and symptoms, interpreted as the result of cardiac ischemia, or abruptly decreased blood flow to the heart muscle. The subtypes of ACS include Unstable Angina (UA) and Myocardial Infarction (MI). Acute MI is the single most common cause of death for both men and women in the developed world. Several data mining studies have analyzed different types of patient data in order to generate models that are able to predict the severity of an ACS. Such models could be used as a basis for choosing an appropriate form of treatment. In most cases, the data is based on electrocardiograms (ECGs). In this preliminary study, we analyze a unique ACS database, featuring 28 variables, including: chronic conditions, risk factors, and laboratory results as well as classifications into MI and UA. We evaluate different types of feature selection and apply supervised learning algorithms to a subset of the data. The experimental results are promising, indicating that this type of data could indeed be used to generate accurate models for ACS severity prediction.