Humboldt-Universität zu Berlin - Mathematisch-Naturwissenschaftliche Fakultät - Wissensmanagement in der Bioinformatik

Humboldt-Universität zu Berlin | Mathematisch-Naturwissenschaftliche Fakultät | Institut für Informatik | Wissensmanagement in der Bioinformatik | Research | Publications | 2003 | A statistical approach to analyse positional dependencies in protein domains: Supplementary information

A statistical approach to analyse positional dependencies in protein domains: Supplementary information

A statistical approach to analyse positional dependencies in protein domains

Jörg Hakenberg1,+, Hubert Hug2,$, and Rainer Schuler1,*

1 Abt. Theoretische Informatik, Universität Ulm, D-89069 Ulm, Germany
2 Universitätskinderklinik, Prittwitzstraße 43, D-89075 Ulm, Germany
+ New address: Jörg Hakenberg, Dept. Knowledgemanagement in Bioinformatics, Humboldt-University Berlin, Rudower-Chaussee 25, D-12489 Berlin, Germany.
$ New address: Hubert Hug, Head of Molecular Medicine and Pharmacogenomics, TheraSTrat AG, Gewerbestrasse 25, CH-4123 Allschwil, Switzerland.
* Corresponding author: Rainer Schuler, Abt. Theoretische Informatik, Universität Ulm, D-89069 Ulm, Germany.
 Phone: +49.731.5024107, Fax: +49.731.5024102, email: schuler@informatik.uni-ulm.de


Abstract

Motivation: Protein domains are usually identifed by conserved regions in the primary amino acid structure. Protein functions are dependent on the 3-dimensional structure and biochemical properties of the amino acids. 3-Dimensional structures are stabilized by interactions between amino acids which are not necessarily adjacent in the primary structure. To preserve the structure, even if certain residues of the sequence are mutated, similar interactions must be possible. This property must be reflected by positional dependencies between residues of positions taking part in an interaction. We use a statistical approach to identify such positional dependicies in aligned protein sequences of the death domain family and the homeodomain.

Results: We identifed a 12 amino acid region with high correlations in the death domain and in the CARD but not in the DED. Most positions with known mutations in death domain family members showed a strong correlation with other positions in the protein. The correlated positions identifed in the homeodomain were distributed over the whole sequence. Point mutations in the positions of the homeodomain leading to a defect were either correlated or conserved, i.e. showed a low entropy.

Contact: schuler@informatik.uni-ulm.de


Please see http://theorie.informatik.uni-ulm.de/Bioinformatics/ for further information.