OBSOLETE Portable Annotation Rules

Daniel Haft, JCVI; 2008

The JCVI is developing a collection of mixed-evidence annotation rules, under the working name BrainGrab/RuleBase (BGRB). A rule has two parts. The first is the set of conditions that must be met for the rule to fire. The second is the set actions to be taken for rules that have fired. BGRB rules are designed to serve as proxies for the annotators that create them. They have very high fidelity but may have low coverage. Types of evidence used in combination include HMM hits and BLAST matches, hits to neighboring genes, pathway reconstruction reports from the Genome Properties system, and species taxonomy. BLAST matches are described by a number of separate parameters for raw score, percent sequence identity, and coverage of total sequence length by the match region. These parameters are customized for each protein family in order to achieve high fidelity in automated annotation systems. The flexible syntax makes it possible to use existing protein family classifiers, such as Pfam and TIGRFAMs HMMs, in new ways. It is especially useful in assigning GO terms to proteins such as SelD (selenide, water dikinase) that have different roles in different contexts.