From Protein Structure to Function with Bioinformatics.pdf

5 Structure and Function of Intrinsically Disordered Proteins 117Table 5.1 Disorder predictors. The table lists the most often used disorder predictors, their URLaddresses, and the principles they are based on. Further details on the predictors are found in thetext, and in references (Ferron et al. 2006; Dosztanyi et al. 2007)Predictor URL PrinciplePONDR VSL2 http://www.ist.temple.edu/disprot/ SVM a with non-linear kernelpredictorVSL2.phpDISOPRED2 http://bioinf.cs.ucl.ac.uk/disopred SVM, NN b for smoothingIUPred http://iupred.enzim.hu Estimated pairwise interaction energyDisEMBL http://dis.embl.de Neural networkGlobPlot http://globplot.embl.de Amino acid propensity, preference forordered secondary structureFoldUnfold http://skuld.protres.ru/~mlobanov/ Amino acid propensityogu/ogu.cgiFoldIndex http://bip.weizmann.ac.il/fldbin/ Amino acid propensityfindexNORSphttp://cubic.bioc.columbia.edu/ Secondary structure propensityPreLinkaSupport vector machinebneural networkservices/NORSphttp://genomics.eu.org/spip/PreLinkAmino acid propensity + hydrophobiccluster analysis5.3.2 Charge-Hydropathy PlotThe classical approach to assess the disordered status of a protein is based on theobservation of Uversky that a combination of low mean hydrophobicity and highnet charge distinguishes IDPs from ordered proteins. This principle can be appliedin a simple fashion, by plotting net charge vs. net hydrophobicity (Uversky et al.2000), in a plot termed either charge-hydropathy (CH) plot or Uversky plot. On theplot IDPs tend to be positioned in the high net charge – low net hydrophobicityregion, and are separated from globular proteins by a linear function of a formula< charge > = 2.743 < hydropathy > − 1.109 (Fig. 5.1), determined at high precisionin a later study (Oldfield et al. 2005a). A limitation of the CH plot is that it onlyenables a binary classification of proteins, without providing information at aminoacid resolution. To deal with this situation, Sussman and colleagues have extendedthis principle (Prilusky et al. 2005) by applying a sliding window along a proteinsequence to calculate mean hydrophobicity and net charge and thereby predict thedisorder of the middle residue (FoldIndex, Fig. 5.2).5.3.3 Propensity-Based PredictorsConceptually related to these approaches are other propensity-based predictors,which assess if a given disorder-related amino acid feature is enriched or depletedwithin a pre-defined segment of the protein. GlobPlot (Linding et al. 2003a),