Protein Evolutionary Relationships from Structure Similarities

One step towards a unifying theory of protein evolution is the reconstruction of protein relationships based on their structural similarity. There have been several efforts made in quantifying structural similarities between proteins.4'43'44 The ambiguity in all of these efforts arises from complications in rigorous quantitative definition of structural similarity. Semi-intuitive definitions of folds have been employed to construct two popular databases, SCOP4 and CATH.43The main drawback of these databases is that they are somewhat subjective.

The FSSP database based on the DALI structure comparison algorithm44 defines a quantitative measure of structural similarity, the Z-score. However, selection of the threshold value Zmm of the Z-score, beyond which proteins are considered structurally similar, also introduces an element of ambiguity into FSSP-based family classification. In a recent paper,45 Getz and coauthors provided a quantitative relationship between FSSP, CATH and SCOP classifications. These authors noted that the matrix of pairwise Z-scores can be viewed as a weighted graph, where each two proteins that have similarity Z > 2 (Z = 2 is the minimal Z-score reported in FSSP) are connected by an edge that carries weight corresponding to the Z-score similarity between these two proteins. Getz et al45 employed clustering algorithms, developed for weighted graphs, to identify fold families. However, clustering of weighted graphs is not exact as it may depend on the chosen algorithm and other factors. Another well-known problem with structural classification of whole proteins presented in FSSP is so-called "floats" where two structurally unrelated proteins having a common "promiscuous" domain are identified as structurally similar. It is, therefore, crucial to reconstruct protein structural relationships taking into account the problem of "floats".

* The divergent evolution and prevalence of more designable structures do not contradict each other.12

0 0

Post a comment