Interaction Profile-Based Protein Classification of Death Domain

Peer reviewed: 
Yes, item is peer reviewed.
Scholarly level: 
Faculty/Staff
Final version published as: 

BMC Bioinformatics 2004, 5:75 doi:10.1186/1471-2105-5-75

Date created: 
2004
Abstract: 

Background: The increasing number of protein sequences and 3D structure obtained fromgenomic initiatives is leading many of us to focus on proteomics, and to dedicate our experimentaland computational efforts on the creation and analysis of information derived from 3D structure.In particular, the high-throughput generation of protein-protein interaction data from a feworganisms makes such an approach very important towards understanding the molecularrecognition that make-up the entire protein-protein interaction network. Since the generation ofsequences, and experimental protein-protein interactions increases faster than the 3D structuredetermination of protein complexes, there is tremendous interest in developing in silico methodsthat generate such structure for prediction and classification purposes. In this study we focused onclassifying protein family members based on their protein-protein interaction distinctiveness.Structure-based classification of protein-protein interfaces has been described initially by Ponstinglet al. [1] and more recently by Valdar et al. [2] and Mintseris et al. [3], from complex structures thathave been solved experimentally. However, little has been done on protein classification based onthe prediction of protein-protein complexes obtained from homology modeling and dockingsimulation.Results: We have developed an in silico classification system entitled HODOCO (Homologymodeling, Docking and Classification Oracle), in which protein Residue Potential InteractionProfiles (RPIPS) are used to summarize protein-protein interaction characteristics. This systemapplied to a dataset of 64 proteins of the death domain superfamily was used to classify eachmember into its proper subfamily. Two classification methods were attempted, heuristic andsupport vector machine learning. Both methods were tested with a 5-fold cross-validation. Theheuristic approach yielded a 61% average accuracy, while the machine learning approach yielded an89% average accuracy.Conclusion: We have confirmed the reliability and potential value of classifying proteins via theirpredicted interactions. Our results are in the same range of accuracy as other studies that classifyprotein-protein interactions from 3D complex structure obtained experimentally. While ourclassification scheme does not take directly into account sequence information our results are inagreement with functional and sequence based classification of death domain family members.

Language: 
English
Document type: 
Article
Rights: 
You are free to copy, distribute and transmit this work under the following conditions: You must give attribution to the work (but not in any way that suggests that the author endorses you or your use of the work); You may not use this work for commercial purposes; You may not alter, transform, or build upon this work. Any further uses require the permission of the rights holder (or author if no rights holder is listed). These rights are based on the Creative Commons Attribution-NonCommercial-NoDerivatives License.
Statistics: