We propose fragment-based methods to estimate and reduce potential function errors in large biomolecular systems. We have generated databases comprising probability density functions describing likely potential function errors for different classes of biomolecular fragment interactions. By analyzing the molecular contacts present in a large system such as a protein or protein-ligand complex, one can estimate the error contributions from each fragment. These error estimates are propagated over all fragments to generate an error estimate for the composite system. Systematic error can be removed, and random error can be reported as an error bar. (Intramolecular) basis set superposition error can be estimated with a statistics-based model. Single-microstate error estimates can be propagated throughout statistical ensembles, yielding error estimates in free energy. In our analysis we observed that increased local sampling of potential energy surfaces naturally reduces potential modeling errors. We suggest general error-handling protocols for biomolecular modeling with imprecise potential functions.

Fragment-based methods for error estimation in biomolecular modeling

245th National Meeting (2013)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.