We found a significant negative correlation between expression rate and the effective number of codons, Ncp, in O. tauri (fig. 2 A, n = 5,463, Spearman ρ = −0.22, P < 10−), CUB increasing with gene expression.
An alternative hypothesis to explain the positive relationship between expression rate and CUB in organisms with an excess of GC optimal codons could be a transcription-coupled AT→GC mutation rate (see Steele [2009] for a review on transcription coupled mutation processes).
The relationship between expression rates and gene lengths in our data set is tenuous (Spearman ρ = 0.05, P < 10−) and has no confounding effect on the relationship between Ncp and expression rates (Spearman partial correlation taking gene length into account ρ = −0.22, P < 10−).
We used the R pcor.test function from Kim and Yi (2006) to estimate partial correlation coefficients between expression rates, CUB, and gene length.
To further substantiate the importance of NQO1 expression in cervical SCC progression, we analyzed the correlation between the high-level NQO1 expression rate and clinical stages of cervical SCCs.
The relationship between expression rates of CDC2 and pathologic grades of human gliomas was analyzed with correlation coefficient.
We thus conclude that the agreement between Figures 1 and 3 is due to the increased correlation between the normalized expression levels for each gene relative to that present in the raw data, Figures 1 and 2. In summary, an explicit form of length bias between expression rates measured by microarray and abundance within the EST database methods has been found.
Further, the amount of data available makes most models except linear ones behind reach, and one therefore assumes a linear relation between the expression rates and the expression levels [15] [17] (or some non-linear transformation of them, the equations still being linear, though [18]).
The null hypothesis of neutral evolution of CUB, as a consequence of mutational bias, does not predict a positive relationship between gene expression rates and CUB, unless the expression process itself is mutagenic.
Here, we first investigate whether we can find evidence of translational selection on CUB in O streococcus tauri and in its virus OtV5 (Derelle et al. 2008) by analyzing the relationship between CUB, expression rates, and tRNA gene prevalence both in the host and in the viral genome.
