Abstract
BACKGROUND: Several in silico methods exist that were developed to predict protein interactions from the copious amount of genomic and proteomic data. One of these methods is Domain Fusion, which has proven to be effective in predicting functional links between proteins. RESULTS: Analyzing the structures of multi-domain single-chain peptides, we found that domain pairs located less than 30 residues apart on a chain are almost certain to share a physical interface. The majority of these interactions are also conserved across separate chains. We make use of this observation to improve domain fusion based protein interaction predictions, and demonstrate this by implementing it on a set of Saccharomyces cerevisiae proteins. CONCLUSION: We show that existing structural data supports the domain fusion hypothesis. Empirical information from structural data also enables us to refine and assess domain fusion based protein interaction predictions. These interactions can then be integrated with downstream biochemical and genetic assays to generate more reliable protein interaction data sets.