We present and test two methods to use quantum chemical calculations to improve standard protein structure refinement by molecular dynamics simulations restrained to experimental NMR data. In the first, we replace the molecular mechanics force field (employed in standard refinement to supplement experimental data) for a site of interest by quantum chemical calculations. This way, we obtain an accurate description of the site, even if a molecular-mechanics force field does not exist for this site, or if there is little experimental information about the site. Moreover, the site may change its bonding during the refinement, which often is the case for metal sites. The second method is to extract a molecular mechanics potential for the site of interest from a quantum chemical geometry optimisation and frequency calculation. We apply both methods to the two Ca2+ sites in the epidermal growth factor-like domains 3 and 4 in the vitamin K-dependent protein S and compare them to various methods to treat these sites in standard refinement. We show that both methods perform well and have their advantages and disadvantages. We also show that the glutamate Ca2+ ligand is unlikely to bind in a bidentate mode, in contrast to the crystal structure of an EGF domain of factor IX.
Download full-text PDF |
Source |
---|---|
http://dx.doi.org/10.1007/s10858-004-6729-7 | DOI Listing |
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!