Detecting periodic patterns in biological sequences.

Abstract

The search for repeated patterns in DNA and protein sequences is important in sequence analysis. The rapid increase in available sequences, in particular from large-scale genome sequencing projects, makes it relevant to develop sensitive automatic methods for the identification of repeats. A new method for finding periodic patterns in biological sequences is presented. The method is based on evolutionary distance and 'phase shifts' corresponding to insertions and deletions. A given sequence is aligned to itself in a certain sense, trying to minimize a distance to periodicity. Relationships between different such periodicity measures are discussed. An iterative algorithm is used, and the running time is nearly proportional to the sequence length. The alignment produces a periodic consensus pattern. A 'phase score' is used to indicate a statistical significance of the periodicity. Three examples using both DNA and protein sequences illustrate how the method can be used to find patterns. On request from the authors. evindc@mat nu.no; finn.drablos@unimed.sintef.no