The present article applies unsupervised machine learning approach to identify patterns of dependence in data of biomarkers, cognitive and demographic characteristics useful for the diagnosis and treatment planning of Alzheimer disease. Two important questions are aimed to reveal in the research– what is the natural structure of the data and whether the groups found are representative for the disease status diagnosis. The answer of these questions is found by applying well established datamining step procedure. This goal is aimed at supporting new methods for diagnosing and predicting Alzheimer’s disease.