Документ взят из кэша поисковой машины. Адрес оригинального документа : http://mccmb.belozersky.msu.ru/2013/abstracts/abstracts/23.pdf Дата изменения: Thu Mar 7 20:01:34 2013 Дата индексирования: Thu Feb 27 20:41:59 2014 Кодировка:

Protein 3D architectures and structural motifs: automated annotation E.A. Aksianov1, A.V. Alexeevski
1 2

1,2

Belozersky Institute, Lomonosov Moscow State University, Leninskie Gori 1, Moscow, Russia Scientific Research Institute for System Studies (NIISI RAN), Moscow, Russia evaksianov@gmail.com, aba@belozersky.msu.ru

Description of a protein 3D structure in terms of structural domains, architecture, topology and structural motifs is a hard problem. Indeed, there are continuous transitions between, for example, -sandwiches and open -barrels, and other pairs of folds; there is no standard means to describe arrangements of -sheets and -helixes in 3D space; minor structural elements (short -helixes, -hairpins, etc.) and uncertain secondary structure assignments complicate formal fold descriptions. These obstacles lead to significant discrepancies in classification of the same structures in popular databases SCOP and CATH [1]. Existing automatic tools for fold description (TOPS, PTGL, PROMOTIF and others) mainly rely on a protein topology (a scheme of contacts of consecutive -strands and helixes) rather than architecture (spatial arrangement of -sheets and -helixes). In contrast, an expert examining a new structure perceives first an architecture of a protein; it takes more his efforts to recover topology. Here we present a detector of the architecture of input protein 3D structure. In its algorithm we attempted to follow expert's decision making. The algorithm is implemented as ProtOn web-service (http://mouse.belozersky.msu.ru/proton). ProtOn consists of 3 main programs. SheeP [2] is a detector of -sheets. It enhances -sheets recovered from secondary structure assignments made by DSSP or STRIDE detectors in order to obtain refined -sheets more adequate to human judgments. SheeP presents -sheets by sheet maps (Fig. 1B). ArchiP is a detector of an architecture of a protein structure. It was designed for detection of architectures of all- and / classes only. ArchiP identifies architectural units and constructs a graph of their contacts (Fig. 1C). Units are of types: -barrels (which not necessarily coincide with whole sheet), not barrel -sheets, parts of -sheets, layers of -helixes (-layers) adjacent to the units of -type. In the graph edges corresponding to main and side contacts are distinguished.


A core of an architecture is detected as a connected component with respect to the main edges only. Units that contacts with the core are considered as extension of the architecture. A B C D

Fig. 1. Example of -sandwich, PDB code 1CWU, chain A. A. Cartoon model. Small -sheet is shown in blue. B. SheeP output: maps of -sheets. Cells of the tables correspond to residues, rows correspond to strands. Residues located on the same side of the -sheet are colored similarly. C. ArchiP output: architectural graph. Boxes correspond to -sheets. All -units are included in appropriate boxes. Vertices (ovals) outside boxes correspond to -layers. A side of a -unit contact is shown by a triangle at the end of corresponding edge. Main contacts are shown by double edges. Thus, core of the architecture consists of the -sheet (unit_4) and two -layers (unit_0 and unit_2) contacting with the -sheet from opposite sides. The core is extended by small -sheet (unit_3). D. MotAn output: ne of detected -motifs.

Architectural graphs of a dozen of common architectures (-sandwichs, -barrels, -propellers, TIM-barrels etc.) are of special type. If the graph of core units coincides with one of these graphs, then ArchiP describes the architecture in common terms. In other cases the output is architectural graph only. To check ArchiP we have detected architectures in all SCOP domains of 125 folds that have exact architectural annotations. Correct architecture assignment were obtained 74% of domains, from 66% for -propellers to 99% for /-sandwiches. The 3rd program, MotAn, detects structural motifs, namely, interlocks (87% correct assignments), jelly-rolls (86%), -helixes (99%), -motifs (99%), meanders (80%). We checked that MotAn assignments are the same or substantially better than assignments of PTGL [3],another structural motif detector, for motifs common for both programs. The work was partially supported by RFBR grants 10-07-00685-, 11-04-91340-. 1. G.Csaba et al. (2009) Systematic comparison of SCOP and CATH: a new gold standard for protein structure analysis, BMC Structural Biology, 9:23. 2. E.Aksianov, A.Alexeevski (2012) SheeP: A Tool for Description of -sheets in Protein 3D Structures, JBCB, 10:2. 3. P.May et al. (2010) PTGL: a database for secondary structure-based protein topologies, NAR, 38:D326-D330