Table 6

The ability of each set of ESEs to predict trends in relative synonymous codon usage
Dataset Negatives Positives P binomial Rho P corr Chi2 P overall
ESR 54 33 0.016 -0.21 0.056 14.04 <0.001
INT2 46 41 0.33 -0.18 0.1 6.82 <0.05
INT2.400 57 30 0.0025 -0.14 0.2 15.2 <0.0005
INT3 56 31 0.0048 -0.24 0.027 17.90 <0.0005
INT3.400 60 27 0.00026 -0.18 0.1 21.11 <0.0001
Ke-ESE 23 64 1 0.0015 0.99 0.02 ns
Ke-ESE400 23 64 1 -0.091 0.4 1.83 ns
PESE 49 38 0.14 0.00033 1 3.93 ns
RESCUE 66 21 7.10E-07 -0.31 0.0031 39.9 <0.0000001

Here was ask whether: (a) each ESE dataset can predict which of two synonymous codons is preferred near a boundary and which is relatively preferred in ESEs, assayed by their HPI scores; and (b) whether the extent of the difference in tendency to be found in ESEs predicts the degree of difference in the preference as one approaches exon ends. Regarding the first aspect, the expectation is that, orientating all comparisons such that the difference in HPI >0, the difference in slope should be negative. We thus ask whether there are more negative values than positives under a directional binomial test. As regards issue (2), we expect a negative correlation: a codon strongly preferred in ESE should be relatively strongly enriched near a boundary, hence a big difference in the slope of the codon usage near the boundary. We compute an overall P value combining the P values of the two tests using Fisher’s method to generate a chi squared value, with 2 degrees of freedom. Those indicated in bold are significant after Bonferonni correction (P <0.05/9).

Cáceres and Hurst

Cáceres and Hurst Genome Biology 2013 14:R143   doi:10.1186/gb-2013-14-12-r143

Open Data