The paper deals with a distance based multi-relational clustering application in a real data case study. A novel method for a dissimilarity matrix calculation in multi-relational settings has been proposed and implemented in R language. The proposed method has been tested by analyzing public actions related to data mining subject and indexed in the medical index database MedLine. Clustering based on partitioning around methods was used for the semi-automated identification of the most popular topics among the MedLine publications. The algorithm implements greedy approach and is suitable for small data sets with a limited number of 1:n relational joins.