Recognition of an organism from fragments of its complete genome

This paper considers the problem of matching a fragment to an organism using its complete genome. Our method is based on the probability measure representation of a genome. We first demonstrate that these probability measures can be modeled as recurrent iterated function systems (RIFS) consisting of four contractive similarities. Our hypothesis is that the multifractal characteristics of the probability measure of a complete genome, as captured by the RIFS, is preserved in its reasonably long fragments. We compute the RIFS of fragments of various lengths and random starting points, and compare with that of the original sequence for recognition using the Euclidean distance. A demonstration on five randomly selected organisms supports the above hypothesis.