PAST: fast structure-based searching in the PDB

Abstract

PAST is a new web service providing fast structural queries of the Protein Data Bank. The search engine is based on an adaptation of the generalized suffix tree and relies on a translation- and rotation-invariant representation of the protein backbone. The search procedure is completely independent of the amino acid sequence of the polypeptide chains. The web service works best with, but is not necessarily limited to, shorter fragments such as functional motifs—a task that most other tools do not perform well. Usual query times are in the order of seconds, allowing a truly interactive use. Unlike most established tools, PAST does not prefilter the dataset or exclude parts of the search space based on statistical reasoning. The server is freely available at .

Proteins

“…47 In this respect structure is a useful indicator of function; indeed most known protein folds are associated with a particular function or functional milieu. 18 Programs that scan the Protein Data Bank (PDB) for structural similarity given a query sequence include, among others, FATCAT, 48 PAST 49 and VAST. 50 However, knowledge of 3D protein structure alone is not always sufficient to accurately infer function.…”

Prediction of Protein Functions

“…In this respect structure is a useful indicator of function; indeed most known protein folds are associated with a particular function or functional milieu (7). Programs that scan the Protein Data Bank (PDB) for structural similarity given a query sequence include, among others, FATCAT (41), PAST (42), and VAST (43). However, knowledge of 3D protein structure alone is not always sufficient to accurately infer function.…”

Ph2

“…PAST [15] is a web-based tool for searching for local structural motifs. The user provides a PDB file or extract from a PDB file as input, and the tool searches for other PDB files that have similar, local structures.…”

“…PH2 was run on the cluster of the IBM/Google Cloud Computing University Initiative. The cluster has 419 dual processor 2.8GHz Xeon machines (15,4), each with 1MB of L2 cache. On this cluster, there is no direct control over the number of physical processors used: this is determined by the number of data files (and in principle, system load -although we have been able to mitigate this by only running when the system is not loaded).…”