From Protein Structure to Function with Bioinformatics.pdf

154 B.H. Dessailly and C.A. Orengo6.3.3 Function Divergence During Protein EvolutionThe traditional approach for annotating a protein of unknown function is to look forhomologies between that protein and other well-characterised proteins, and to transferthe functional annotations from the latter to the former, assuming that proteins thatdescend from a common ancestor should share some degree of common functionality(Whisstock and Lesk 2003). But it is now a well-established fact that this approach iserror-prone and that its incautious application results in unmanageable propagation oferroneous annotations in databases (Devos and Valencia 2001).The major source of errors in this process is that the assumption following whichhomologous proteins have similar functions is inaccurate (Devos and Valencia2000). There are now numerous documented cases of related proteins with verydifferent functions, including the long-known example of hen egg-white lysozymeand mammalian α-lactalbumin that share more than 35% sequence identity andhave very similar structures. Yet, it is reasonable to assume that the larger the evolutionarydistance between two homologous proteins, the lower the probability ofthese proteins sharing the same function. Several studies have attempted to determinesequence identity cut-offs that would safely guarantee conservation of functionbetween pairs of homologues, but results are somewhat discordant and theissue is still under debate (Todd et al. 2001; Rost 2002; Tian and Skolnick 2003;Sangar et al. 2007). One likely explanation for the difficulty to derive universalsequence identity cut-offs for reliable transfer of function annotations betweenhomologues is the above-mentioned fact that different superfamilies have very differentpatterns of sequence and function divergence. Accordingly, many recentstudies focus on the analysis of sequence-structure-function relationships in specificsuperfamilies or subsets thereof, and may reveal highly valuable insights as tohow the variations in sequence and structure correlate with variations in function.In the following section, we describe function variation within superfamilies inmore detail, with particular emphasis on the mechanisms thought to bring aboutthis variation.6.3.3.1 Function Diversity at the Superfamily LevelThe sequences of proteins classified in the same superfamily have sometimesdiverged beyond levels that can be detected by standard sequence alignment methods.Even though three-dimensional structures are generally accepted to be farmore conserved than sequences during evolution, major differences can still beobserved between the structures of remote homologues. Such structural differences canarise from insertions/deletions (indels) of large elements of secondary structures oreven several of these. A recent study of indels amongst homologous structuresshowed that it is not uncommon for successive insertions of secondary structures tooccur in the same location of the fold of a protein during evolution, thus giving riseto so-called nested indels (Jiang and Blouin 2007). Another analysis of insertions