Abstract

With the increasing number of protein structures[11] in Protein Data Bank(PDB)[1] efficient and accurate algorithms similar to local alignment of sequences[5] are necessary to classify newly discovered structures into appropriate Super Families,Folds and Classes[4][2]. This problem was partially addressed by some of the Structural Alignment algorithms[9][10][8][7] which can be used to align the new structures to the existing structures in the PDB and classify the new structure. We have identified two major drawbacks (Alignment at side chain level, Local Structural Alignment) which none of the existing state of the art structural alignment algorithms [9][10][8][3] ever addressed. In this paper we provide algorithms which can align proteins considering both side chain and backbone conformations, also we address the problem of Local Structural Alignment similar to Local Sequence Alignment[12], Local Structural Alignment can be used in identification of Structural Motifs. Since we consider side-chain conformations our algorithms are more accurate compared with existing algorithms[3]. We validate the accuracy of our method taking SCOP[2] as golden standard, our algorithm acheives an average Super Family accuracy of 84.09% and Class accuracy of 86.93% whereas algorithms like PSIST[3] could only acheive 68.18% , 73.29% of Super Family and Class accuracy. We also validate our Local Structural Alignment by searching for Tyrosine phosphorylated substrates [14] in some of the PDB structures and identify the functional structural motif in 1C86,1LAR,2GJT,2NV5,2H4V(PDB ID's).