In this study, we choose steric clashes, SASA/MSA, void volume and percentage of unsatisfied hydrogen-bond donor/acceptors as metrics reporting on the quality of packing of the protein core and the formation of proper contacts in the core and shell. We measure the quality of any given structure in terms of these metrics by comparing against the benchmarks from high-resolution crystal structures.
Servers like MolProbity and WHAT IF (Davis et al., 2007
; Hooft et al., 1996
; Vriend and Sander, 1993
) have revolutionized structural biology by providing accurate assessment of the quality of a structure through evaluation of clashes, hydrogen bonds and protein covalent geometry. Through Gaia, we seek to complement already existing servers. Gaia is unique in several aspects of protein structure quality assessment. For example, Gaia
provides a unique way to define clashes using energetics compared to fixed overlap distance cutoffs (0.4 Å) used by other servers. Furthermore, while other servers provide a list of hydrogen bonds, Gaia also evaluates unsatisfied hydrogen bond partners in a protein. Finally, Gaia
uses a novel method to compute void volume in a protein and provides a statistical score for the total void volume of a protein. Thus, Gaia
provides a systematic, multi-faceted evaluation of the quality of a protein structure model, including clashes, voids, hydrogen bonds and molecular surface, in addition to the local, covalent geometry of individual residues and peptide bonds.
In addition to introducing tools for quantifying the quality of protein structures, our benchmarks revealed interesting properties of protein cores. Our finding that majority of small and medium sized proteins feature no voids, suggests that single domains are well packed in proteins, while folding of multiple domains may introduce voids. Further analysis is required to substantiate this hypothesis. The distribution of the percent buried polar atoms forming no hydrogen bonds peaks at zero reiterating the strong penalty for the burial of unsatisfied polar atoms. We also observe a high probability for voids that can fit just one water molecule, implying a prevalence of singly occurring buried water that may be important for structural stability. The possible structural role of buried waters is further supported by the observation of substantial number of contacts between buried polar atoms and crystallographic waters in the structures in our dataset. This discovery is important in modeling protein structure given that most current methods for ab initio protein structure prediction do not consider buried structural waters.
The filters established in this study combined with calculation of covalent geometry of proteins are available for use by protein-structural biology community as a web server (Gaia
). For a given input protein, Gaia
calculates all the above properties and provides P
-values by comparing the input protein parameters to the distributions obtained from high-resolution crystal structures. By providing a detailed report on each of these properties within minutes, Gaia
serves as a final filter for estimating the quality of a given protein structural model.
Funding: American Heart Association Predoctoral Fellowship (to S.R. 09PRE2090068); the University of North Carolina Research Council (to F.D.); the National Institutes of Health (to N.V.D. grant number R01GM080742 and ARRA supplements GM080742-03S1 and GM066940-06S1).
Conflict of Interest: none declared.