From Protein Structure to Function with Bioinformatics.pdf

152 B.H. Dessailly and C.A. Orengo6.3.1.2 Practical DefinitionsOnly databases that consider structural data are described here.CATH and Gene3D – In the CATH classification, domains in a given topology (seeSection 2.1.2) are further classified in the same Homologous superfamily (H-level) ifthey are believed to have a common ancestor. Two domains are considered homologousif they satisfy at least two of the following criteria: (a) structural similarity, assessedusing empirically-derived cut-offs; (b) sequence similarity, assessed using standardsequence comparison methods and HMM sequence searches; and (c) functional similarity,identified using manual analysis. Gene3D expands this classification to proteinsof unknown structure, by scanning sequences against a library of CATH profile-HMM’s, thus matching parts of these sequences to CATH homologous superfamilies(Yeats et al. 2008). CATH superfamilies are further divided into sequence families thatare defined at different cut-offs of sequence identity. A cut-off of 35% sequence identityis used to define non-redundant groups of proteins (s35 families).SCOP and Superfamily – For SCOP superfamilies, homologies are determinedby sequence similarity or by manual comparison of structural and functional features(Andreeva et al. 2008). This manual assignment provides the community witha curated expert classification of domain structures, but suffers from the concomitantdrawback that any manual process is inevitably prone to subjective decisions.Domains are classified into the same SCOP family if they are “clearly evolutionarilyrelated”. In practise, this definition generally means that protein domains aregrouped into the same family if they share pair wise residue identities of more than30%. However, some domains are classified into the same SCOP families in theabsence of high sequence identities if similar structures and functions providedefinitive evidence of common ancestry. This has the advantage of allowing forsome flexibility in the assignment of homology relationships but also gives moreroom for subjectivity in the process. The Superfamily database expands SCOP toproteins of unknown structure by annotating sequences with SCOP descriptions atthe family and superfamily level (Wilson et al. 2007). As with Gene3D, Superfamilyuses SCOP-based HMM profiles to assign matches in sequences.SFLD – The Structure-Function Linkage Database has been developed morerecently with the specific aim of studying the structure-function relationships amongsthomologous enzymes. It currently covers a relatively small set of superfamilies, ascompared with CATH and SCOP, but provides a detailed description of the evolutionof function within these superfamilies. The SFLD imposes that enzymes within superfamiliesshould not only be homologous but must share a mechanistic attribute in thecatalytic reaction using conserved structural elements (Pegg et al. 2006). SFLD familiesconsist of enzymes that perform the same overall reaction in a given superfamily.6.3.2 Evolution of Protein SuperfamiliesUltimately, the criterion to group proteins together in superfamilies is that the genesencoding them descend from a common ancestor gene. The processes by which anancestor gene gives rise to two (or more) copies of itself are commonly referred tounder the term duplication.