Figure 3.

Effect of overall amount of bias on behaviour of codon usage measures. Figure 3a describes the change in behaviour of each measure as the overall bias increases from unbiased ('None') to a nucleotide composition noted on the x axis. The values were obtained from 10000 randomly generated sequences, 2500 codons long, per frequency table (None, Low-1, Low-2 etc.) per measure. Figure 3b demonstrates how the measures react when the nucleotide compositions of the generated sequences and the expected codon frequency table are interchanged (commutative property). See Results and Discussion for further explanation. In both figures, the values on the y axis are expressed as percentages of the measures' dynamic ranges.