Protein structure cross validation

One of the most useful and widely accepted cross validation methods is the use of Rfree (Brunger, 1992). A subset of reflections are withheld from the minimization process and only checked periodically for their agreement to Fc. If Rfree drops during the refinement, the crystallographer obtains an unbiased indication that the refinement has gone well and no gross errors have been introduced in the model. 

In the sections that follow, the algorithms of PROCHECK, WHATCHECK, VerifySD, ERRAT, and PROVE will be described individually, and program outputs illustrated and interpreted for the particular case, PDB ID code lja3 (Dimasi et al., 2002). After surveying all the stmctures deposited in the PDB, lja3 was flagged 

1 Backbone lepiesentation of the model of lja3 as deposited with the PDB (black) and after revision based on the output of the various cross validation programs (grey). The region in the upper third of the figure (residues 204-230) required the most rebuilding. 

The stmcture validation programs described here can be accessed individually (web addresses given in the References section at the back of this book), or conveniently all together using the Structure Analysis and Validation Server (SAVS) at http //nihserver.mbi.ucla.edu/SAVS. 

The PROCHECK stmcture validation program (Laskowski et al, 1993) deploys a multitude of quality checks using both aoss-validation and standard checks for geometric deviations. The Ramachandran plot it produces is its most attractive and useful feature (Ramakrishnan and Ramachandran, 1965). The plot consists of a two dimensional plot of phi/psi values for each amino acid residue. Steric overlap between side chain and main chain atoms limits the energetically allowed values 

---

Another historical method is the NOE cross-validation.47 This is performed by a partial elimination of NOEs from a full set of NOEs. The elimination is random. Calculated structures from the different sets of partially eliminated NOEs are statistically analyzed in comparison with the structure calculated from the full set of NOEs. The predictability of the eliminated NOEs from remaining NOEs is assessed as structural consistency (and therefore high quality) of the eliminated NOEs. With increased computing power, the statistical analysis will be feasible to determine high-resolution protein structure. 

To cross-validate the results, each group of structurally related proteins is left out of the training set in turn and used to test the network. Such a partitioning scheme (in contrast to a jackknife one, for example) minimizes the likelihood of biasing the results in favor of structural descriptors (see Section II). Its use yields true predictions (denoted cv ) in contrast to fits of the data, in which aU the proteins are included during the training (denoted tm )- The latter tend to yield inflated accuracy statistics, but we describe them here as well for comparison with earlier studies [12,13,20,47], which failed to cross-validate their results [however, it should be noted that the relationship in Ref. 12 has been used successfully for blind predictions (K. W. Plaxco and D. Baker, personal communication)]. 

---