A note on independence assumption on binding sites in biological sequence analysis

Finding significant patterns from sequence data is an important issue in various fields extending well beyond biology. Despite their potential utility, a difficulty in direct application of the recent methods stems from the independence assumption between sites. Markovian assumptions have recently been made for either informative binding sites or non-informative background sites. In this article, it is shown that under certain specifiable conditions, making the independence assumption yields better results in finding binding sites from sequence data with substantial Markovian dependency.