Codon usage bias is due to a combination of biased mutation and selection effects. Isochore-related GC mutational bias has been shown to be the dominant cause of tissue-specific codon bias in adult human tissues and limited evidence for translational selection has been shown. This thesis is a comprehensive evaluation of the relative contribution of selection and mutation to variation in codon usage among different mouse embryonic tissues. Through a detailed analysis of tissue-specific codon

more »

... e in relation to gene expression and gene location of mouse embryonic tissue-specific genes, I found that translational selection is partially responsible for some differences in codon usage among tissue-specific genes. This observation indicates profound impact of selection favouring codon-anticodon adaptation. The characteristic GC biases of tissue-specific gene sets are shown not to be caused by their clustering on the same isochore. Tissue-specific genes are no more clustered on the genome than randomly selected genes. In mouse, the usage of some G-ending codons decreases with increasing GC bias, while all other G and C-ending codons increase. To understand this counterintuitive observation, we generate a continuous-time Markov chain model of GC-biased synonymous substitution which explains qualitative usage patterns of all codons, including non-linear and sometimes non-monotone codon usage in isoleucine, arginine and leucine. This effect is universal, extending to mouse and human genes as well as plant and prokaryotic genomes. This work enriched our understanding of the codon-anticodon adaptation theory and extended it to the level of tissue-specific genes. The result suggests the possibility of tissue-specific tRNA pools mediating tissue-specific codon-anticodon adaptation. The methods developed in the thesis can be easily extended to characterize this previously little explored facet of codon-anticodon adaptation.