부산대학교 도서관

The use of local covalent geometry for quality assessment and refinement of protein structure models is a well-established methodology. The question arises whether information on non-covalent geometry contained within resolved structures can be harnessed to improve structure prediction. Moreover, incorporation of different combinations of priors would pave the way towards multi-body potentials. Existing empirical force-fields do not facilitate an interactive exploration of the parameter space and an assignment of spatial propensities to contacts. Hence, we investigate the possibility of making such propensities available for synergistic modeling. We present an approach that facilitates the extraction and analysis of anisotropic contact potentials for a multitude of parameters describing an amino acid and the conditions within its microenvi-ronment. For this purpose, two novel visualization principles will be introduced. The first visualization illustrates anisotropic residue-dependent contact density potentials in the form of a map projection. A second visualization is overlaid onto this, showing similar local neighborhoods as abstract traces of residues contained within each individual neighborhood. The Contact Geometry Analysis Plugin (CGAP) (for CMView) we developed allows incorporation of geometric orientation propensities into the process of interactive protein modeling and can be used for the generation of improved energy functions. It further supports the analysis of model quality, as it directly illustrates model consistency with known spatial propensities which, in turn, enables users to detect possible structural errors.