Simplicial edge representation of protein structures and alpha contact potential with confidence measure

Xiang Li,Changyu Hu,Jie Liang

DOI: https://doi.org/10.48550/arXiv.physics/0302082

2003-02-23

Abstract:Protein representation and potential function are essential ingredients for studying proteins folding and protein prediction. We introduce a novel geometric representation of contact interactions using the edge simplices from alpha shape of protein structure. This representation can eliminate implausible neighbors not in physical contact, and can avoid spurious contact between two residues when a third residue is between them. We develop statistical alpha contact potential. A studentized bootstrap method is then introduced for assessing the 95% confidence intervals for each of the 210 parameters. We found with confidence that there is significant long range propensity (>30 residues apart) for hydrophobic interactions. We test alpha contact potential for native structure discrimination using several decoy sets, and found it often has comparable performance with atom-based potentials requiring more parameters. We also show that alpha contact potential has better performance than potential defined by cut-off distance between geometric centers of side chains. Clustering of alpha contact potentials reveals natural grouping of residues. To explore the relationship between shape representation and physicochemical representation, we test the minimum alphabet size for structure discrimination. We found that there is no significant difference in discrimination when alphabet size varies from 7 to 20, if geometry is represented accurately by alpha simplicial edges. This result suggests that the geometry of packing plays an important role, but the specific residue types are often interchangeable.

Biological Physics,Quantitative Biology