Singling out functional similarities in graph databases