The similarity between the sequences of the i-th fragments of two promoters (i=1,..7) was assumed non-random if a higher alignment score was obtained for them using the program FASTA than for random sequences with the same oligonucleotide content in 95% of cases. The measure of similarity for the group as a whole was the ratio of the number of non-randomly similar pairs of sequences to the total number of all possible pairs in the group.
As turned out, the degree of similarity for the promoters of a species just insignificantly exceeded the random threshold, the highest similarity being observed in the TATA-box and its vicinity. As to the tissue-specific promoters, it was typical of them to have on the whole higher degrees of similarity within a species. That the promoters of non-homologous genes contain non-randomly similar fragments uncovers an important role of convergence in the evolution of promoters.