Maximal words in sequence comparisons based on subword composition

Abstract

Measures of sequence similarity and distance based more or less explicitly on subword composition are attracting an increasing interest driven by intensive applications such as massive document classification and genome-wide molecular taxonomy. A uniform character of such measures is in some underlying notion of relative compressibility, whereby two similar sequences are expected to share a larger number of common substrings than two distant ones. This paper reviews some of the approaches to sequence comparison based on subword composition and suggests that their common denominator may ultimately reside in special classes of subwords, the nature of which resonates in interesting ways with the structure of popular subword trees and graphs. © Springer-Verlag Berlin Heidelberg 2010.