Using Neighborhood Preserving Projections for Comparing Protein Structures
Sourangshu Bhattacharya, Chiranjib Bhattacharyya
IISc-CSA-TR-2004-14
(November 2004) Available formats: [pdf]
Filed on November 29, 2004
This paper reports a new protein structure comparison algorithm that uses a
novel spectral graph-matching technique. We view protein structures as weighted
graphs. The problem of comparing protein
structures is posed as a weighted graph matching problem - a problem which is
also studied in fields like Computer Vision. A novel similarity
measure between the vertices of the two graphs is proposed based on the spectralproperties of the graph. Finally, a new technique for retrieving the correspondences
between the residues of the two proteins based on the new similarity function is
developed. The program is tested to detect the correct correspondences in case
of structurally similar protein pairs. Moreover, our program gives the full
structural alignment even in the case when the protein sequences are related by
circular permutations, which are not detected by many standard protein
structure comparison programs.
Please bookmark this technical report as http://aditya.csa.iisc.ernet.in/TR/2004/14/.Problems ? Contact techrep@csa.iisc.ernet.in
[Updated at 2009-10-22T06:42Z]