The vast number of expressed sequence tags (ESTs) in public databases provides an important resource for comparative and functional genomics. A variety of methods based on homology search or domain profile search have been developed to functionally annotate protein domains in ESTs. However, these methods either ignore potentially valuable information from the homologues beyond the top N hits, or they are extremely time consuming. We provide an efficient and novel tool, called E2D (EST to Domain), which functionally annotates anonymous ESTs by recognizing potential domains from the enlarged hit proteins. Comparison with InterProScan shows that E2D is more efficient and effective for domain recognition. Additionally, we achieve 87.5% agreement with existing GO function annotations in TIGR through domain-GO mapping, which demonstrates the efficacy of our approach. E2D is available at http://bc03.iis.sinica.edu.tw/LEE/E2D.htm.