Computational Prediction of Domain-domain Interactions: Factor-graph Based Modelling and Inference

Proteins interact with each other to perform an array of cellular functions. A systems-level knowledge of protein interaction networks, as well as the underlying domain-domain interactions, constitutes a major goal in systems biology. In this mini-review, we discuss some background to the problem of inference of domain-domain interactions and briefly discuss some relevant computational and statistical approaches. Specifically, we review our earlier work in which we formulate this problem using a graphical model combined with inference of domain-interaction probabilities using powerful statistical inference methods, namely belief propagation. The inferred interactions can be used for analysing network data, and, by looking at the specificity of protein-protein interactions, they can also be used to predict novel protein interactions within and across orthologous species.