Optimization and Analysis of Weighted-Window Predictors of Structural Disorder in Proteins
X-ray crystallographic protein structures often contain disordered regions that are observed as missing electron density. We have developed single sequence and profile-based weighted-window predictors of structural disorder in proteins, as well as a simple method for addressing disorder-prone chain termini in disorder prediction. Optimizing the parameters for these relatively simple predictors with crystallographic data using a simulated annealing type algorithm, we achieve performance similar to that of DISOPRED2. Optimized parameters from these disorder predictors provide information relating to physical processes underlying crystallographic disorder. Optimized score adjustment values suggest a simple, monotonic relationship between disorder and residue distance from termini that is nearly the same for amino- and carboxy-terminal positions. Residue disorder parameters are strongly associated with scales from certain experimental model systems that primarily reflect hydrophobic interactions. Our data do not suggest a strong association between crystallographic disorder and secondary structure beyond that explained by hydrophobicity. Our results lend support to the idea that while hydrophobic side chain interactions are primarily involved in determining stability of the folded conformation, hydrogen bonding and similar polar interactions are primarily involved in conformational and interaction specificity.