Substitutional Tolerant Markov Models for Relative Compression of DNA Sequences

摘要

Referential compression is one of the fundamental operations for storing and analyzing DNA data. The models that incorporate relative compression, a special case of referential compression, are being steadily improved, namely those which are based on Markov models. In this paper, we propose a new model, the substitutional tolerant Markov model (STMM), which can be used in cooperation with regular Markov models to improve compression efficiency. We assessed its impact on synthetic and real DNA sequences, showing a substantial improvement in compression, while only slightly increasing the computation time. In particular, it shows high efficiency in modeling species that have split less than 40 million years ago.