Mining Protein Contact Maps

The 3D conformation of a protein may be compactly represented in a symmetrical, square, boolean matrix of pairwise, inter-residue contacts, or "contact map". The contact map provides a host of use- ful information about the protein's structure. In this paper we de- scribe how data mining can be used to extract valuable information from contact maps. For example, clusters of contacts represent cer- tain secondary structures, and also capture non-local interactions, giving clues to the tertiary structure. In this paper we focus on two main tasks: 1) Given the database of protein sequences, discover an extensive set of non-local (fre- quent) dense patterns in their contact maps, and compile a library of such non-local interactions. 2) Cluster these patterns based on their similarities and evaluate the clustering quality. We show via experiments that our techniques are effective in characterizing con- tact patterns across different proteins, and can be used to improve contact map prediction for unknown proteins as well as to learn protein folding pathways.