Efficient computation of codon bias spectra and their implications in genome sequences

Codon Bias Spectra (CBS) locate the regions in genome sequences where the distribution of synonymous codons are biased and non-uniform. In this paper, we show that polyphase filterbank concepts are applicable for CBS computation, and analyze CBS patterns in genomes such as C. glutamicum and C. elegans. This method provides a parallel and efficient way for CBS calculation along with its accessories like spectra of G+C or A+T combinations. Patterns obtained in CBS provide information for gene expressions and evolutionary studies. Correlation and regression analysis are performed here for CBS patterns to estimate the preponderance of synonymous codons for a particular amino acid. Our analysis show the higher correlation exists for GC3 spectra that AT3 spectra and regression line show biased selection of synonymous codon for a amino acid. We also show sensitivity-specificity analysis for C. elegans genome for statistical validation.