Figure 2
The effect of randomizing synonymous codon identity at different positions, on the prediction accuracy of the synonymous codon identity at the central position. The relative codon assignment accuracy is the comparison between the accuracy using true and random synonymous codons and it is expected that this ratio will be 1 if true and random synonymous codons have equal prediction power, as suggested by the null hypothesis. Mean and \(95\%\) confidence intervals were estimated using bootstrapping: the original test set of size \(N_{\alpha }=9575\) and \(N_{\beta }=3378\) was subsampled with replacement to form 10, 000 independent samples of size \(M_{\alpha }=6000\) and \(M_{\beta }=3000\) for the \(\alpha\) and \(\beta\) mode respectively. The corresponding p values for rejecting the null hypothesis are listed in parenthesis, and the calculation of the p values is explained in greater detail in the “Methods”.
