Is there a close relationship between synonymous codon bias and codon-anticodon binding strength in human genes?

Abstract

Synonymous codon bias has been examined in 78 human genes (19967 codons) and measured by relative synonymous codon usage (RSCU). Relative frequencies of all kinds of dinucleotides in 2,3 or 3,4 codon positions have been calculated, and codon-anticodon binding strength has been estimated by the stacking energies of codon-anticodon bases in Watson-Crick pairs. The data show common features in synonymous codon bias for all codon families in human genes: all C-ending codons, which possess the strongest codon-anticodon binding energies, are the most favored codons in almost all codon families, and those codons with medium codon-anticodon binding energies are avoided. Data analysis suggests that besides isochore and genome signature, codon-anticodon binding strength may be closely related to synonymous codon choice in human genes. The join-effect of these factors on human genes results in the common features in codon bias.