12.07.2015 Views

From Protein Structure to Function with Bioinformatics.pdf

From Protein Structure to Function with Bioinformatics.pdf

From Protein Structure to Function with Bioinformatics.pdf

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

208 E.C. Meng et al.proteins sharing a GO molecular function classification, and proteins sharing bothSCOP superfamily and GO molecular function classifications. Redundancy-filteredsets of structures were used. A motif was generated from each member structureof a positive group, and members of the remaining groups were used as the negativeexamples. Motifs were only generated for groups <strong>with</strong> at least six structures afterredundancy filtering. The server uses RIGOR (Kleywegt 1999) for searching.Potential commercial users must first contact RIGOR’s author for licensing information(see the Uppsala Software Fac<strong>to</strong>ry web site, Table 8.2). Statistical significanceis estimated <strong>with</strong> the function developed by the authors of PINTS (Stark et al. 2003).8.4 Related MethodsHybrid point-surface and single-point-centred descriptions of local structure cannotbe strictly categorized as 3D motif approaches, but share many similarities.Methods primarily based on surface descriptions are covered in Chapter 7.8.4.1 Hybrid (Point-Surface) DescriptionsCavbase (Schmitt et al. 2002; Kuhn et al. 2006) and SiteEngine (Shulman-Peleg et al.2004) describe binding sites as collections of pseudoa<strong>to</strong>ms and their associatedsurface patches. The pseudoa<strong>to</strong>ms represent surface-exposed functional groups ofvarious types, such as hydrogen bond donor or accep<strong>to</strong>r. Comparisons involve findinggeometrically and physicochemically consistent sets of pseudoa<strong>to</strong>ms, superimposingstructures based on those matches, and then scoring based on surface patchoverlap and physicochemical similarity. Surface points typically far outnumber thepseudoa<strong>to</strong>ms, so scoring is relatively computationally demanding. The two methodsdiffer in details of the pseudoa<strong>to</strong>m types and scoring, and in how matching isperformed: Cavbase performs clique detection, while SiteEngine uses geometrichashing. Cavbase and an associated database of sites are available as part of thecommercial software package Relibase+ licensed by the Cambridge CrystallographicData Centre. The SiteEngine web server (Shulman-Peleg et al. 2005) (Table 8.3)performs pairwise comparisons but not database searches. A SiteEngine executablefor Linux can be downloaded for noncommercial use only (Table 8.3).8.4.2 Single-Point-Centred DescriptionsThe program FEATURE (Bagley and Altman 1995) describes local structure as aset of properties in concentric shells emanating from a single point. The propertiesinclude descrip<strong>to</strong>rs of a<strong>to</strong>ms, functional groups, residues, secondary structure, and

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!