On the precision of calculated solvent-accessible surface areas

The fact that protein structures are dynamic by nature and that structure models determined by X-ray crystallography, electron microscopy (EM) and nuclear magnetic resonance (NMR) spectroscopy have limited accuracy limits the precision with which derived properties can be reported. Here, the issue of the precision of calculated solvent-accessible surface areas (ASAs) is addressed. A number of protein structures of different sizes were selected and the effect of random shifts applied to the atomic coordinates on ASA values was investigated. Standard deviations of the ASA calculations were found to range from ∼10 to ∼80 Ų. Similar values are obtained for a handful of cases in the Protein Data Bank (PDB) where `ensembles' of crystal structures were refined against the same data set. The ASA values for 69 hen egg-white lysozyme structures were calculated and a standard deviation of the ASA of 81 Ų was obtained (the average ASA value was 6571 Ų). The calculated ASA values do not show any correlation with factors such as resolution or overall temperature factors. A molecular-dynamics (MD) trajectory of lysozyme was also analysed. The ASA values during the simulation covered a range of more than 800 Ų. If different programs are used, the ASA values obtained for one small protein show a spread of almost 600 Ų. These results suggest that in most cases reporting ASA values with a precision better than 10 Ų is probably not realistic and a precision of 50–100 Ų would seem prudent. The precision of buried surface-area calculations for complexes is also discussed.