Computational Approaches to Predict Protein–protein and Domain–domain Interactions

Knowledge of protein and domain interactions provides crucial insights into their functions within a cell. Various high throughput experimental techniques such as mass spectrometry, yeast two hybrid, and tandem affinity purification have [Q3] generated a significant amount of large-scale high throughput protein interaction data [9,19,21,28,29,35,36,58]. Advances in experimental techniques are paralleled by the rapid development of computational approaches designed to detect protein– protein interactions [11,15,24,37,45,46,48,50]. These approaches complement experimental techniques and, if proven to be successful in predicting interactions, provide insights into principles governing protein interactions. A variety of biological information (such as amino acid sequences, coding DNA sequences, three-dimensional structures, gene expression, codon usage, etc.) is used by computational methods to arrive at interaction predictions. Most methods rely on statistically significant biological properties observed among interacting proteins/domains. Some of the widely used properties include co-occurence, coevolution, co-expression, and co-localization of interacting proteins/domains.