Auto-selection order of Markov chain for background sequences with chi-square test

摘要

Modeling non-coding background sequences appropriately is important for the detection of regulatory elements from DNA sequences. Based on the chi-square statistic test, some explanations about why to choose higher-order Markov chain model and how to automatically select the proper order are given in this paper. The chi-square test is first run on synthetic data sets to show that it can efficiently find the proper order of Markov chain. Using chi-square test, distinct higher order context dependences inherent in ten sets of sequences of yeast S. cerevisiae from other literature have been found. So the Markov chain with higher-order would be more suitable for modeling the non-coding background sequences than an independent model.