From Protein Structure to Function with Bioinformatics.pdf

118 P. TompaFig. 5.1 Charge-hydropathy plot of protein disorder. Net charge vs. mean hydrophobicity hasbeen plotted for intrinsically disordered (full diamond) and ordered (empty circle) proteins. Thetwo are separated by a straight line < charge > = 2.743 < hydropathy > − 1.109, with arrows pointingto the lines delimiting the zone with a prediction accuracy of 95% for disordered proteins and97% of ordered proteins, at the expense of discarding 50% of all proteins (Reprinted withpermission from Oldfield 2005a. Copyright 2005 American Chemical Society)for example, applies a measure based on a scale, expressing the propensity for agiven amino acid to be in a region of random coil vs. a regular secondary structure.DisEMBL (Linding et al. 2003b) uses three optional features in a similar fashion,“loops/coils” (in accord with the DSSP classification), “hot loops” (loops with highB-factors in X-ray crystal structure), and “Remark 465” that describes the propensityof an amino acid to be missing from the PDB X-ray structures.A slightly different approach is applied for propensity-based prediction byPreLink (Coeytaux and Poupon 2005). PreLink combines the two properties ofdisordered regions (defined as linker regions connecting globular domains) thatthey have a biased amino acid composition and they usually contain no or smallhydrophobic clusters. To quantify these two properties, amino acids distributions inordered proteins and disordered regions have been calculated. In the querysequence, the distance to the nearest hydrophobic cluster is computed by an automatedHydrophobic Cluster Analysis (HCA), which is based on a two-dimensional(2D) helical representation of protein sequences. Disorder score is based on bothamino acid composition and this distance.5.3.4 Predictors Based on the Lack of Secondary StructureA different, although not entirely unrelated approach relies on the propensity ofamino acids to form secondary structural elements (α-helix, β-strand and coil) withthe underlying assumption that long regions (>70 consecutive amino acids) devoid