Web Portal on Computational Biology
Note: Articles considered especially worth reading are highlighted using this color .
Computational methods for the prediction of protein interactions
Valencia et al. (2002): Computational methods for the prediction of protein interactions
This paper briefly describes "the five computational techniques available for the prediction of interaction partners": "presence or absence of genes in related species", "conservation of gene neighborhood", "gene fusion evenets"; "similarity of phylogenetic trees (mirrortree)" and "in silico two-hybrid method". The authors "examine their range of applicability" and "analyze new trends in the determination of interacting surfaces on the basis of sequence information". A chapter that compares these methods with each other follows:
Unfortunately, a definitive evaluation of any of these methods cannot yet be undertaken, because the availability of collections of interacting proteins is still highly limited. [...] Complementary to these efforts [to develop databases of protein interactions and to establish standards for the exchange of information between these databases], various data-mining procedures are emerging for the automatic extraction of information about protein interactions from the vast amount of accumulated bibliographic information.
Then, a chapter about the "prediction of the molecular basis of protein interaction" follows:
[A] set of new computational methods can now address the problem of the prediction of interacting surfaces in the absence of complete information about the corresponding structures of the binding proteins. Initial approaches have been based on the observed properties of the statistical composition of interacting surfaces in terms of residue types (polarity, charge, etc.) and on the structure of the surfaces. [...] A second type of method addresses the prediction of interacting residues in the absence of structural information. The first reported application determines the distribution of positions that show family-dependent patterns of conservation in MSAs ( tree determinants ). [...] A promising alternative to that described above is the use of information about correlated mutations in order to highlight the interaction sites in binding proteins.
The authors conclude:
The combination of experimental and theoretical data could, for the first time, provide complete information about interaction networks, thereby allowing studies to be undertaken of the distribution and number of interactions, the presence of key nodes in the networks, tolerance to perturbations and differences in network organization from one organism to another.
Contact: cdvolko (at) gmail (dot) com
Imprint: This website is owned by Claus Volko, Vienna, Austria. No liability is taken for the contents of any of the linked websites. http://www.cdvolko.net/