The Emergence of Pattern Discovery Techniques in Computaional Biology

Recent years have witnessed an emergence of pattern discovery methodologies for solving numerous tasks which arise in computational biology. Known also as ``data mining'' approaches, they represent a novel approach for extracting useful information from databases containing various types of biological information. Initially, dynamic programming techniques were applied to the analysis of biological sequences and to the determination of sequence similarity between a query sequence and one or more biological sequences (DNA, proteins, and fragments) from a collection. Subsequent studies of such sequence similarity revealed conserved functional and structural signals, thus making the argument for the usefulness of such approaches. Research effort spanning almost two decades gave rise to a number of useful algorithms and an abundance of interesting scientific results [2, 58, 63, 77]. Almost in parallel, researchers began looking into other approaches in an effort to develop concise consensus sequences that captured and represented regions of similarity across several sequences presumed to be related. A large number of early methods relied on multiple string alignment [19, 25, 44, 70] as the method of choice for discovering these regions [22, 54, 58, 60, 80, 89, 91]. The related sequences could be transformed to one another through permissible edit operations (e.g., mutations, insertions, deletions) each of which had an associated cost. doi:10.1006 mben.2000.0151, available online at http: www.idealibrary.com