Mining Over-Represented 3D Patterns of Secondary Structures in Proteins

ABSTRACT We consider the problem of ̄nding over-represented arrange- ments of Secondary Structure Elements (SSEs) in a given dataset of representative protein structures. While most pa- pers in the literature study the distribution of geometrical properties, in particular angles and distances, between pairs of interacting SSEs, in this paper we focus on the distribu- tion of angles of all quartets of SSEs and on the extraction of over-represented angular patterns. We propose a variant of the Apriori method that obtains over-represented arrange- ments of quartets of SSEs by combining arrangements of triplets of SSEs. This speci ̄c case will pose the basis for a natural extension of the problem to any given number of SSEs. We analyze the results of our method on a dataset of 300 non redundant proteins