A Novel Protein Soft Docking Algorithm Based on Spectral Analysis of Shape Complementarity

We describe a novel method for predicting protein docking conformations using spectral analysis techniques. Rigid-body docking methods based on shape complementarity, particularly Fourier transform based techniques are successful for docking monomers with conformations adopted at binding (complexed conformations) [4]. This approach is very efficient because a complete search over all six degrees of freedom can be performed by simple rotation and translation operations. However, when docking unbound conformations, the method generally yields many false positive structures. Several rigid-body methods have been developed to overcome this problem [1, 2, 3, 5, 6]. Mandell et al. employed an energy term consisting of both the sum of Poisson-Bolzmann electrostatic and van der Waals energies [5]. Palma et al. introduced the concept of soft-surface rigid-body docking introducing molecular flexibility in the rigid-body algorithm. Amongst the protein complexes observed, ARG, LYS, ASP, GLU and MET present the highest frequency and amplitude of movements between the structures of free and co-crystallized proteins. So these amino acids are considered flexible and are allowed to unrealistically penetrate the other molecule during the docking search. The “soft-surface” representation is achieved by simply setting a 0 in the core matrix cells assigned to the side chain atoms of these five amino acid types [6]. Here, we propose a genuine soft docking algorithm that uses spectral analysis for handling shape complementrarity in a more ample way without constraining side chains by type of amino acid. This algorithm extracts a characteristic shape from the unbound conformation by applying a three dimensional filter. The high flexibility of the amino acids on the molecular surface is averaged and thus the filtering introduces “softness” in the algorithm. The softened shapes prevent unrealistic penetrations into one another of the molecules during the docking search process. The algorithm has been implemented as a module within the system for macromolecular interaction assessment MIAX[2], the results outperforming those of flexible docking both in MIAX and as compared to other similar systems.