Bacteria DNA sequence compression using a mixture of finite-context models

The ability of finite-context models for compressing DNA sequences has been demonstrated on some recent works. In this paper, we further explore this line, proposing a compression method based on eight finite-context models, with orders from two to sixteen, whose probabilities are averaged using weights calculated through a recursive procedure. The method was tested on a total of 2,338 sequences belonging to bacterial genomes, with sizes ranging from 1,286 to 13,033,779 bases, showing better compression results than the state-of-the-art XM DNA coding algorithm and also faster operation.