Isaac Newton Institute - Abstracts
INI Home Page

Analysing and predicting contacts between domains in interacting proteins -- (undated)
Thomas Nye, (MRC Biostatistics Unit, Cambridge)
Proteins frequently bind together in pairs or larger complexes to take part in cellular processes or biochemical events. Recently there has been considerable interest in identifying domains that occur frequently in interacting proteins. This information is often used to predict whether two proteins will interact on the basis of the domains they contain. However, such predictions are fraught with difficulties given the limitations of available interaction data. Here we address an alternative problem: given two proteins that are already known to interact and lists of the domains they contain, can we predict the pattern of contacts between the constituent domains? Predictions of this kind are of biological value, since the exact structure of a protein complex has critical implications for its function. A maximum likelihood method is used to estimate a propensity of interaction between pairs of domains given an experimentally observed set of interactions. For each domain pair the propensity represents the probability that the two domains come into contact when they are present inside an interacting protein pair. The propensities are then used to predict domain contacts within protein pairs that are known to interact. The Protein Data Bank (PDB) contains the full 3-dimensional structures for a limited set of interacting protein pairs. This information is used to extract the pattern of domain contacts for these interactions. Our predictive method is validated against these contacts.