Protein Structure and Function: Applications of Bioinformatics Methods - John Rigden 2014

Bioinformatics Methods for Studying the Structure and Function of Disordered Proteins
Prediction of IDP Functions
Combining Sequence and Disorder Information: Phosphorylation Sites and CaM-Binding Motifs

There are two known cases where the accuracy of predicting short characteristic motifs in Proteins was improved by incorporating information on structural disorder, specifically regarding phosphorylation sites and calmodulin-binding domains (CaMBTs). By comparing experimentally verified phosphorylation sites with potential sites that are typically unphosphorylated (at Ser, Thr, or Tyr residues), Dunker and colleagues (Iakoucheva et al. 2004) found that regions adjacent to phosphorylation sites are enriched in amino acid residues that promote structural disorder and depleted in residues that favor order (Dunker et al. 2001). By combining positive and negative training datasets while accounting for local disorder, it is possible to develop a method for predicting phosphorylation sites. The DISPHOS algorithm (disorder-enhanced phosphorylation predictor) demonstrates enhanced prediction accuracy compared to other phosphorylation site prediction tools, such as NetPhos (Blom et al. 1999) and Scansite (Obenauer et al. 2003).

Another well-characterized example is the interaction between calmodulin (CaM) and its target molecules, which involves significant flexibility on the part of both interacting partners. It is known that CaM typically wraps around its target molecule (CaMBT), a helical peptide approximately 20 Amino Acids long (Ikura and Ames 2006). Comprehensive analysis revealed that target disorder is a prerequisite for recognition by CaM (Radivojac et al. 2006). For instance, CaM-dependent Enzymes are often activated by partial proteolytic Cleavage—such as calcineurin (Manalan and Klee 1983) or cyclic nucleotide phosphodiesterase (Tucker et al. 1981)—suggesting the presence of local disorder within the binding regions. Accounting for structural disorder has been successfully applied to develop a CaMBT prediction method with improved performance (Radivojac et al. 2006).