Currently there are over 12000 Structural Genomics (SG) protein structures deposited in the protein data bank, of which over 6000 have been assigned a putative function based on a global structural or sequence match and over 3000 are of unknown function. Such annotations of SG proteins according to the closest structure or sequence match very often lead to incorrect assignment of a putative function. These misannotations then propagate in the databases, as additional proteins are annotated by automated means.We have developed a new computational approach to predict protein function reliably and to identify these misannotations. The method, called Structurally Aligned Local Sites of Activity (SALSA), starts with sets of functionally important residues, predicted with computed electrostatic and chemical properties. These are sets of spatially localized amino acid residues that participate in the biochemical function. These localized sets of residues are then structurally aligned to identify what proteins of the same function have in common in the local regions of their active sites. Predicted residues with the same relative positions in proteins of common function then constitute a chemical signature for that functional type. Protein structures of unknown function are then analyzed and compared to the chemical signatures to predict biochemical function. Application to the members of the six-hairpin glycosidase superfamily is presented.  Using this method, we show that a putative alpha-rhamnosidase from Bacteroides thetaiotaomicron is likely to be correctly annotated, while a putative glucoamylase from Bacteroides fragilis is likely to be incorrectly annotated.

Functional classification of six-hairpin glycosidase superfamily with structurally aligned local sites of activity

247th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.