UCSF

All enzymes can be classified into superfamilies that share common mechanistic attributes of catalysis. These catalytic attributes are delivered via variants of a common structural scaffold. The superfamily context, through the linkage of structure and aspects of function, is useful for making inferences about the function of uncharacterized proteins. However, the superfamily context is more than a few isolated model proteins--it is encoded in a landscape of proteins that evince the full range of variation of structure and function across all members of the superfamily. Protein similarity networks are a projection of this landscape.

Similarity networks were validated and then used in combination with other analyses to create an atlas of the thioredoxin fold that exposes the heterogeneity of the glutaredoxins, and to show how catalytic residues are distributed throughout the class. An important change in the activation of the coenzyme of some glutathione transferases was described, paired with a new example of convergent evolution of the coenzyme binding site. Additionally, a similarity landscape was used to provide context for recent developments in the biology of cysteine peptidases in parasites. Most of the proteins in all of these classes have not been functionally characterized, and this is one of the first attempts to examine their catalytic capabilities. The similarity landscapes of the thioredoxin fold, glutathione transferases, and cysteine peptidases can be used to prioritize and guide functional characterization of the many proteins that remain unexamined.