Dicodon-based measures improve the prediction performance of gene expressivity

Motivation: Codon usage preference patterns have been associated with modulation of translation
efficiency, protein folding and mRNA decay. However, new studies support that codon pair usage has
also a remarkable effect at gene expression level. Here, we expand the concept of CAI to answer if codon
pair usage patterns can be understood in terms of codon usage bias, or if they offer new information
regarding coding translation efficiency.
Results: Through the implementation of a weighting strategy to consider the dicodon contributions, we
have improved the prediction of gene expression level in about 20% in relation to CAI. Interestingly,
we have noted that dicodons associated with low value of adaptiveness are related to dicodons which
mediate strong translational inhibition in yeast. We have also noticed that some codon-pairs have a
smaller dicodons contribution than estimated by the product of the respective codon contributions.