Motivation: Automated annotation of Expressed Sequence Tags (ESTs) is becoming increasingly important as EST databases continue to grow rapidly. A common approach to annotation is to align the gene fragments against well-documented databases of protein sequences. The sensitivity of the alignment algorithm is key to the success of such methods. Results: This paper introduces a new algorithm, FramePlus, for DNA–protein sequence alignment. The SCOP database was used to develop a general framework for testing the sensitivity of such alignment algorithms when searching large databases. Using this framework, the performance of FramePlus was found to be somewhat better than other algorithms in the presence of moderate and high rates of frameshift errors, and comparable to Translated Searchin the absence of sequencing errors. Availability: The source code for FramePlus and the testing datasets are freely available at ftp.compugen.co.il/pub/research. Contact: raveh@compugen.co.il