Skip to main content
. Author manuscript; available in PMC: 2009 Jul 25.
Published in final edited form as: Cell. 2008 Jul 25;134(2):341–352. doi: 10.1016/j.cell.2008.05.042

Table 1.

Translational accuracy selection quantified by the intragenic association between “optimal” codons and conserved sites.

All genes Top 10% highest expression
Organism Z Odds ratio P(better codon set)a Z Odds ratio P(better codon set) a
E. coli 12.62*** 1.108 0.0062 2.69** 1.105 0.0650
Yeast 22.08*** 1.125 0.0283 8.87*** 1.308 0.0004
Worm 25.54*** 1.135 4.58×10−6 7.66*** 1.351 3.90×10−5
Fly 65.72*** 1.362 2.72×10−7 17.72*** 1.430 1.00×10−6
Mouse 35.37*** 1.180 0.0049 5.86*** 1.144 0.0268
Human 31.97*** 1.160 0.0015 6.20*** 1.171 0.0059
Simulation
 +cost 4.07*** 1.437 0.0255 0.82c 3.000c 0.1319c
 −cost −1.59 0.894 0.8277 −1.01 0.807 0.8514
 +noiseb 4.07*** 1.437 0.0255 0.54 1.180 0.3718
*

= P < 0.05

**

= P < 0.01

***

= P < 0.001; all significance levels after false-discovery-rate correction for multiple testing.

a

P(better codon set) is the fraction of synonymous codon sets yielding a stronger association with conserved sites, as assessed by exhaustive enumeration (see Experimental Procedures). For high-expression genes, significance was assessed relative to 106 random alternative sets.

b

Noise added to expression and Fop measurements changes only the set of genes ranking in the top 10% highest-expressed genes.

c

The codon preference test breaks down for these genes because almost all sites are conserved and most codons are optimal. See Figure 5A for direct measurement of the accuracy gain in highly expressed genes.