Goal: Individual characteristics are determined through a genome consisting of a complex base combination. This base combination is reflected in the k-word profile, which represents the number of consecutive k bases. Therefore, it is important to analyze the genome-specific statistical specificity in the k-word profile to understand the characteristics of the genome. In this paper, we propose a new k-word-based method to analyze genome-specific properties. Methods: We define k-words consisting of the same number of bases as statistically identical k-words. The statistically identical k-words are estimated to appear at a similar frequency by statistical prediction. However, this may not be true in the genome because it is not a random list of bases. The ratio between frequencies of two statistically identical k-words can then be used to investigate the statistical specificity of the genome reflected in the k-word profile. In order to find important ratios representing genomic characteristics, a reference value is calculated that results in a minimum error when classifying data by ratio alone. Finally, we propose a genetic algorithm-based search algorithm to select a minimum set of ratios useful for classification. Results: The proposed method was applied to the full-length sequence of microorganisms for pathogenicity classification. The classification accuracy of the proposed algorithm was similar to that of conventional methods while using only a few features. Conclusions: We proposed a new method to investigate the genome-specific statistical specificity in the k-word profile which can be applied to find important properties of the genome and classify genome sequences.