With the exception of the secondary structure assignment program DSSP, which only employs hydrogen bonding information [ 8], a dozen structure assignments programs using geometric features of local substructures and C α atoms have been proposed [ 3, 9] and programs that employ hydrogen bonds and geometrical restraints have also been applied [ 7].
We compared the approach using rules based on minimal, discerning subsets of local substructures, with the approach of using all rules based on one single local substructure.
Here, we provide a comprehensive analysis of the structure-function relationship in proteins, in which a library of recurring multi-fragment structural motifs called local descriptors of protein structure [25], [26] are used to learn IF-THEN rules [27], [28] that associate combinations of local substructures with specific protein functions.
The relationship between structure and function was modeled using IF-THEN rules [27], [28] where the IF-part of each rule specifies a minimal combination of local substructures discerning a particular protein structure from structures annotated to other GO classes (Figure 1C, D).
For comparison, we randomly sampled 1000 pairs of local substructures matching at least two proteins and where the substructures occurred in at least one of the rules.
We used this framework, as implemented in the ROSETTA rough set system [28] (http://rosetta.lcb.uu.se), for learning IF-THEN rules associating combinations of local substructures of proteins with particular GO classes.
In principle, the method finds the minimal sets of local substructures that discern a particular protein from all other proteins annotated to a different GO class.
For each rule and each matching protein, we computed the average fraction of residues in pairs of local substructure that were in contact (residues common to both of the local substructures were not considered).
This greatly differs from any previously published work in terms of completeness, the use of multi-fragment local substructures, and the fact that the relationship to function was explored using no information about functional sites or sequence patterns.
Make use of local resources.
