On Coarse Grained Representations And the Problem of Protein Backbone Reconstruction
(English)Manuscript (preprint) (Other academic)
Crystallographic protein structures reveal that generically, only two of the Ramachandran angles are flexible. The third Ramachandran angle, all the backbone bond angles, and also all the covalent bond lengths are quite rigid, displaying only insignificant deviations from their optimal values. This empirical observation is among the rationale for the construction of coarse grained force fields where only a subset of the full set of atomic coordinates is utilized as dynamically active variables. The present article addresses in a systematic manner the question, to what extent the various angles andbond lengths can be replaced by their optimal values. In the case of bond lengths, it is found that the optimal values are in practice sufficient. But a coarse graining where a subset of angular variables is replaced by optimal values, commonly yields geometrically incorrect protein structures. There appears to be an inherent numerical instability, which seems to reflect the presence of a positive Liapunov exponent in the iterative reconstruction algorithm. Besides the full and complete set of individual atomic angles, essentially only one numerically stable coarse grained subset of angular variables is found. It consists of variable virtual Cα backbone bond and torsion angles. In combination with fixed, constant valued virtual bond lengths these two angles reproduce the original structure with high precision. The present observations impose strong limitations on the subset of backbone coordinates that can be utilized, even in principle, for the development of coarse grained force fields.
Research subject Physical Biology
IdentifiersURN: urn:nbn:se:uu:diva-199985OAI: oai:DiVA.org:uu-199985DiVA: diva2:621897