From Protein Structure to Function with Bioinformatics.pdf

132 P. Tompafied by that function within one category of transcription factors is rather insensitiveto amino acid changes as long as the above character of the domain is maintained(Hope et al. 1988). On the other hand, mutations that change this character impairtrans-activation function (Gill and Ptashne 1987). Thus, some features apparent atthe level of composition are closely related to function.Such possible general relations have been directly addressed by clustering IDPsin the composition space (Vucetic et al. 2003). The starting point of this analysiswas that disorder predictors trained on one group of proteins often perform poorlyon other groups, which indicates significant differences in sequence propertiesamong disordered proteins. Thus, Dunker and colleagues have clustered 145 IDPsby setting up competition among increasing numbers of predictors, with the criterionof prediction accuracy used to partition individual proteins. Three roughlyequally populated “flavours” of disorder, denoted as V, C, and S, could be identified,with slightly different amino acid compositions: flavour C has more His, Met,and Ala, flavour S has less His, and flavour V has more of the least flexible aminoacids (Cys, Phe, Ile, Tyr), than the other flavours. Flavours appear to have somediscernible functional associations. For example, 9 out of 10 E. coli ribosomal proteinsfall into flavour V. On the other hand, IDPs binding to viral genomic RNA arepractically excluded from flavour V, along with DNA binding proteins. IDPsinvolved in protein-protein interactions are associated with flavours V andS. Notwithstanding the limitations of this analysis, it did clearly suggest that thetype of disorder manifested in composition is related to function, and, pending furtheranalysis, can be tentatively used for prediction purposes.5.6 Limitations of IDP Function PredictionAs should be clear from the foregoing discussions, prediction of the function ofIDPs from sequence is fraught with uncertainties. Analysis of the underlying difficultiesmay be approached from different directions which, nevertheless, areentirely intertwined: rapid evolution of IDPs and sequence independence of theirfunction. In the final section these will be discussed in some detail.5.6.1 Rapid Evolution of IDPsFast evolution of IDPs/IDRs has been directly demonstrated by comparing theamino acid replacement rate of disordered and globular regions in protein families,in which both regions are simultaneously present (Brown et al. 2002). In 26 suchfamilies, statistical measure of variability (average genetic distance) was calculatedby comparing each pair of sequences in multiple alignments. The results showedthat the disordered regions evolve much faster than the ordered region in 19 families,at about the same rate in 5 families, and significantly more slowly only in 2