Protein surface atom neighbourhoods classification

Abstract

The paper presents a classification of the protein surface atom neighbourhoods from the hydrophobicity perspective. Hydrophobicity is the property which is considered around each surface atom. The actual hydrophobicity distribution on the atoms that form an atom's vicinity is replaced by an equivalent hydrophobicity density distribution, computed in a standardized octagonal pattern around the atom. All atoms hydrophobicity densities are clustered using K-means algorithm. A three layers neural network is trained for classification of the atoms vicinities having as many nodes in the output layers as clusters are.