Figure 5

Comparison of uni-directional and bi-directional USM implementations for DNA sequences. The similarity matrices for, respectively, d _f and D values between two portions of E. coli K-12 MG1655 threonine gene A (thrA, genome positions 337–2799) and threonine gene B (thrB, genome positions 2801–3733) are presented. The numbers in the axis identify the position in the gene. Actual values of d _f and D are shown for the framed region on the table to the right. a) The d _f values were obtained by a unidirectional implementation of the USM procedure (Eq. 4). By comparing this figure with a similar analysis reported previously [12] for the same sequences (Fig. 10 of that report) it can be seen that they are nearly indistinguishable, even if the exact values vary. The equivalence between unidirectional USM for n = 2 and CGR highlights the property that CGR is a special case of USM. The fact that the latter can be implemented for any value of n or any number of unique units justifies the Universal naming; b) In this plot the same sequences were compared using bidirectional USM, and, accordingly, generate a matrix of D values (Eq.8, 11). It is clearly apparent, and as already noted for Figure 2c, that D-similarity between any two homologous units is an estimate of the length of the entire homologous segment.