Compressing the human genome using exclusively Markov models

Abstract

Models that rely exclusively on the Markov property, usually known as finite-context models, can model DNA sequences without considering mechanisms that take direct advantage of exact and approximate repeats. These models provide probability estimates that depend on the recent past of the sequence and have been used for data compression. In this paper, we investigate some properties of the finite-context models and we use these properties in order to improve the compression. The results are presented using the human genome as example. © 2011 Springer-Verlag Berlin Heidelberg.