Figure 1 -- Density of phototargets, weighted by their frequency in the E. coli chromosome, and calculated for different G+C contents and for three kinds of random genomes. The weights are as follows: 0.59*f(tt)+0.34*[f(tc)+f(ct)]+0.07*f(cc) (where f(xy) is the frequency of dinucleotide xy in the specified genome). Three models of random genomes are analyzed. In the worst case (solid curve), the genome is the concatenation of a sequence of pyrimidines and a sequence of purines: all pyrimidines are involved in a pyrimidine dinucleotide. In the best case (dotted curve), the genome is an unbroken succession of pyrimidine-purine dinucleotides: no pyrimidine is involved in a pyrimidine dinucleotide. In the "random case" (dashed curve), the frequency of a pyrimidine dinucleotide is the result of chance (f(xy) = f(x)*f(y)).
Figure 2 -- Plot of the mean z-score statistics for intergenic sequences (x-axis) and for coding sequences (y-axis), for each of the four pyrimidine dinucleotides. On each plot, a dot corresponds to the mean of these two statistics in a given prokaryote chromosome. The null x and y axis (dotted lines), and the 5% limits of significance for the standard normal distribution (dashed lines) are plotted as benchmarks. It should be noted note (see Fig 3), that the variability within one chromosome is sometimes as great as that between different chromosomes.
Figure 3 -- Each figure shows the distributions of the z-score in all coding sequences corresponding to each of the three strains of Prochlorococcus marinus. In each figure, the distribution for the MED4 (a high-light adapted strain) is shown as a solid line; the distribution for the SS120 (a low-light adapted strain) is shown as a dashed line, and the distribution for the MIT 9313 (a low-light adapted strain) is shown as a dotted line. The 5% limits of significance for the standard normal distribution (dashed vertical lines) are plotted as benchmarks.