Automatic detection of false annotations via binary property clustering

Computational protein annotation methods occasionally introduce errors. False-positive (FP) errors are annotations that are mistakenly associated with a protein. Such false annotations introduce errors that may spread into databases through similarity with other proteins. Generally, methods used to minimize the chance for FPs result in decreased sensitivity or low throughput. We present a novel protein-clustering method that enables automatic separation of FP from true hits. The method quantifies the biological similarity between pairs of proteins by examining each protein's annotations, and then proceeds by clustering sets of proteins that received similar annotation into biological groups.

Using a test set of all PROSITE signatures that are marked as FPs, we show that the method successfully separates FPs in 69% of the 327 test cases supplied by PROSITE. Furthermore, we constructed an extensive random FP simulation test and show a high degree of success in detecting FP, indicating that the method is not specifically tuned for PROSITE and performs well on larger scales. We also suggest some means of predicting in which cases this approach would be successful.

Automatic detection of FPs may greatly facilitate the manual validation process and increase annotation sensitivity. With the increasing number of automatic annotations, the tendency of biological properties to be clustered, once a biological similarity measure is introduced, may become exceedingly helpful in the development of such automatic methods.