SISYPHUS--structural alignments for proteins with non-trivial relationships

Abstract

With the increasing amount of structural data, the number of homologous protein structures bearing topological irregularities is steadily growing. These include proteins with circular permutations, segment-swapping, context-dependent folding or chameleon sequences that can adopt alternative secondary structures. Their non-trivial structural relationships are readily identified during expert analysis but their automatic identification using the existing computational tools still remains difficult or impossible. Such non-trivial cases of protein relationships are known to pose a problem to multiple alignment algorithms and to impede comparative modeling studies. They support a new emerging concept of evolutionary changeable protein fold, which creates practical difficulties for the hierarchical classifications of protein structures.To facilitate the understanding of, and to provide a comprehensive annotation of proteins with such non-trivial structural relationships we have created SISYPHUS ([Sigmaomeganuphiomicronzeta]--in Greek crafty), a compendium to the SCOP database. The SISYPHUS database contains a collection of manually curated structural alignments and their inter-relationships. The multiple alignments are constructed for protein structural regions that range from oligomeric biological units, or individual domains to fragments of different size. The SISYPHUS multiple alignments are displayed with SPICE, a browser that provides an integrated view of protein sequences, structures and their annotations. The database is available from http://sisyphus.mrc-cpe.cam.ac.uk.

Figure 1

SISYPHUS web-interface. (A) Example of a query result using alignment accession code AL10054814. The output page lists the compiled information for the alignment and provides links to other related alignments through DAG relationships, related SCOP nodes and alignments with common attributes (keywords, GO terms). (B) Visualization of the structure-based sequence alignment in Jalview. (C) Visualization in 3D using SPICE. The left panel shows the alignment in a default mode showing the superimposed structures of the aligned regions only. The middle panel lists the PDB codes and the start and end positions of the structurally equivalent regions. SPICE also provides annotations for PDB, UniProt and Ensembl peptides (right panel). The gray strip indicates the location of the alignment region in the PDB, UniProt and Ensembl peptide sequence.

Figure 2

SPICE view of AL10069117 alignment in a ‘full structure’ mode, displaying all chains in the selected PDB entries. Shown in thick backbone are the aligned segments from the representative structures: AhpD (red, 1knc; chain A), TM1620 (yellow, 1p8c; alignment construct (ALC) made of parts of chains A, C, D and E), TTHA0727 (green, 2cwq; ALC of chains B and C), Atu0492 (bluegreen, 2gmy; chain A) and MTH234 (blue, 2af7; ALC of parts of chains A, C, D and E). (A) Similarity between the trimeric assembly of AhpD and the hexameric assemblies of the CMD family members, TM1620, TTHA0727 and MTH234, is highlighted by viewing them along the common 3-fold axis and turning off the display of the fifth (Atu0492) structure. (B) Similar subunit folds, but different oligomeric assemblies of Atu0492 and AhpD are displayed in approximately the same orientation, with the other structures ‘turned-off’. Note the 2-fold axis of the Atu0492 hexamer, which runs in the vertical direction in the figure plane and coincides with the 2-fold axis of the CMD family hexamers in (A).