Finding and characterising candidate binding sites

We present a method for identifying and representing indentations, cavities and holes in a known protein structure. The identification procedure is performed initially. Once a pocket in the protein has been found, we use a boundary representation technique to represent it. Both the identification and the representation procedures rely upon modifications to the Marching Cubes algorithm of Lorensen and Cline (1987), which can be used to give a piecewise planar representation of a three-dimensional object. The identification of a pocket essentially involves finding a point, (x₀, y₀, z₀) say. The point should be such that a sphere of radius r can he centred there, without touching the centre of any atom in the protein, but there exists R>r such that a sphere of radius R centred at (x₀, y₀, z₀) does touch the centre of an atom. If such a point is found, then it is deemed to be in a pocket. The boundary representation technique involves storing information about a boundary in the structure of a graph. The graph structure of a pocket, potential docking site, can then be compared to the graph structure of a candidate binding substrate using standard graph comparison algorithms. The required input for the algorithm is in the form of a Protein Data Bank (PDB) file