Codon usage variability determines the correlation between proteome and transcriptome fold changes

We identified how much of the variability in the correlation between protein and mRNA concentrations can be attributed to the gene codon frequencies. We propose the hypothesis that the influence of codon frequency is due to the competition of cognate and near-cognate tRNA binding; which in turn is a function of the tRNA concentrations. Transcriptome and proteome data were combined in two analytical steps; first, we used Self-Organizing Maps (SOM) to identify similarities among genes, based on their codon frequencies, grouping them into different clusters; and second, we calculated the variance in the protein mRNA correlation in the sampled genes from each cluster. This procedure is justified within a mathematical framework.With the proposed method we observed that in all the six studied cases most of the variability in the relation protein-transcript could be explained by the variation in codon composition.The integration of large scale transcriptome and proteome data along with genome-wide sequence information can give insights into the molecular mechanisms that control cellular functions. Moreover, formulation of mathematical models, either mechanistic or statistic, to express such molecular mechanisms remains a challenging task to understand system properties [1]. The correlation between mRNA transcripts and their corresponding cognate proteins has been found to be positive, but it is not sufficiently good to predict protein levels based on their cognate transcript [2,3]. If all the mRNAs were translated at a constant rate the correlation between mRNA and protein concentration would be high. The observed lack of correlation is therefore due to the particularities of the translation mechanism. For instance, in yeast 73% of the variance in protein abundance is explained by the translation mechanism and only 27% due to the variations of the mRNA concentration [4,5]. To explain the differences in the responses between protein and transcript levels recent studies atte