Work this year focused on an analysis of the effects of various? changes in the specification of gap costs on the retrieval? accuracy of protein profile search algorithms, with a view? towards improving the retrieval accuracy of PSI-BLAST.? ? In brief, the introduction of position-specific substitution? costs has greatly improved the sensitivity of protein sequence? database search programs: PSI-BLAST and related Hidden Markov? Model (HMM) programs are able to recognize much more distant? sequence relationships than can BLAST. HMM programs may employ? position-specific gap costs as well. By modifying these programs,? we found no improvement from position-specific gap costs that? took account of the specific amino acids that were inserted or? deleted. However, we did find a significant improvement from? gap costs that varied according to their location.? ? Incorporating position-specific gap costs into PSI-BLAST requires? an accurate description of PSI-BLAST statistics under their use.? The most promising path to this goal is the adaptation of the? hybrid alignment scoring method for use with PSI-BLAST.
