Graph mining methods have emerged to address the limitations ofitemset mining algorithms when analyzing structured data. It may thereforeappear counterproductive to employ the latter for mining graph data.Nevertheless, for graph classification tasks, where the focus is on predictiveperformance rather than comprehensibility, the use of itemset mining can be asensible alternative to graph mining algorithms. In this paper, we examine thepros and cons of itemset mining on graph data using 18 medicinal chemistrydatasets, and show that the itemset mining algorithms are not only efficient andreliable on graph classification and regression, but also competitive with thegraph mining algorithms.