From Protein Structure to Function with Bioinformatics.pdf

114 P. Tompaproteins lack a well-defined three-dimensional structure under native, physiologicalconditions, however, challenged the universality of this paradigm (Tompa 2002,2005; Dyson and Wright 2005; Uversky et al. 2005). With the rapid accumulationof data in support of this emerging alternative view of proteins, the need for a reassessmentand extension of the structure-function paradigm became compelling(Wright and Dyson 1999).A range of biophysical techniques, primarily X-ray crystallography, NMR,SAXS and CD, have provided evidence that intrinsically disordered, or unstructured,proteins (IDPs/IUPs) or regions of proteins (IDRs) assume no well-definedconformations, but rather a fluctuating ensemble of alternative structural states(Tompa 2002, 2005; Dyson and Wright2005; Uversky et al. 2005). Superficially,they resemble the denatured states of globular proteins, whereas detailed structuralanalyses suggest that different IDPs may occupy conformational states anywherebetween the fully disordered (random coil) and compact (molten globule) stateswith characteristic distributions of transient secondary and tertiary contacts(Uversky et al. 2000; Uversky 2002). At variance with denatured globular proteins,IDP functions directly stem from the unfolded states, and are mostlyinvolved in regulating processes of signal transduction and gene transcription(Iakoucheva et al. 2002; Ward et al. 2004; Tompa et al. 2006). Functional classificationschemes of IDPs are actually based on whether their function directly stemsfrom disorder, or transient/permanent binding to partner molecules (Dunker et al.2002; Tompa 2002, 2005).Not only are IDPs able to function despite their lack of stable structures,structural disorder actually provides functional advantages in regulatory functions,such as the separation of specificity from binding strength (Wright andDyson 1999), adaptability to various partners (Tompa et al. 2005), increased rateof interaction (Pontius 1993) and frequent involvement in post-translationalmodifications (Iakoucheva et al. 2004). These advantages enable IDPs to fit intounique functional niches, and explain the advance of protein disorder in evolution,with a critical difference in frequency between eukaryotes and prokaryotes(Iakoucheva et al. 2002; Ward et al. 2004; Tompa et al. 2006). The advantagesalso explain a high level of disorder in functionally important regulatory proteins,which also play central roles in disease, such as the prion protein (LopezGarcia et al. 2000), BRCA1 (Mark et al. 2005), tau protein (Schweers et al.1994), p53 (Bell et al. 2002), and α-synuclein (Weinreb et al. 1996). The currentmost complete collection of IDPs, the DisProt database (www.disprot.org), containsabout 500 disordered proteins, mostly observed serendipitously as such(Sickmeier et al. 2007). The application of predictors based on such collection ofproteins, however, suggests that, in the proteomes of metazoa, about 5–15%of proteins are fully disordered, and 30–50% of proteins contain at least onelong disordered region (Iakoucheva et al. 2002; Ward et al. 2004; Tompa et al.2006). To narrow this apparently wide gap in knowledge, a lot of effort is spenton developing bioinformatics algorithms to predict disorder and function fromamino acid sequence. This review focuses on the principles and recent developmentsin this area of IDP research.