Use of homology domains in sequence similarity detection.
We have found the detection of homology domains using a nonlinear similarity score and the DD algorithm to be a useful approach for identifying similarity between sequences and evaluating potential homology. There are several reasons for the success of the method. (i) Homology domains are identified by a rigorous method that guarantees they will be locally optimal. (ii) The relative significance of different homology domains can be directly compared using the nonlinear similarity score. (iii) Different cost matrices can be used in the calculation of the similarity score. (iv) Relatively long sequences can be compared in a single pass as storage requirements are proportional to the shorter of the two sequences being compared. (v) The method has proved to be very sensitive in practice. (vi) The boundaries of authentic regions of homology are accurately identified. (vii) The information required to define a homology domain (its location, size, similarity score, etc.) can be stored in a compact data structure, facilitating the sharing of homology domain data among different software tools. (viii) The method can be applied to similarity searches of the nucleotide and protein sequence data banks. These properties make the identification of homology domains for studying sequence similarity a useful companion to other accepted methods, such as dynamic programming based analyses.