Properties of the word set for estimating similarities between prokaryotic genomes in linguistic approach

Recently, as completely sequenced genomes have been rapidly increasing in number, comparison between whole genome sequences is becoming more important. Linguistic approach is one of the available methods to estimate the similarities between long sequences such as whole genomes[1]. In the method, a word set W is constructed, in which a word is defined as a sequence piece of four letters of nucleotides with fixed length, and the frequency of appearance of each word in W is calculated throughout a genome. The similarity between genomes is estimated by comparing the distributions of the frequencies of appearance calculated for the genomes based on such as the Kendall’s rank correlation. In performing the linguistic approach, we must predetermine three properties of W : the word length L, the size n, and the contents. In the previous study[2], we obtained the result, by analyzing the word diversities in prokaryotic genomes, that L = 8 ∼ 12 is appropriate for prokaryotic species. In this study, we investigate the other two properties, the size and the contents of W adequate for analyzing the similarity between prokaryotic genomes.