Find in Library

In the natural host, most of the synonymous codons of a gene have been evolutionarily selected and related to protein expression and function. However, for the design of a new gene, most of the existing codon optimization tools select the high-frequency-usage codons and neglect the contribution of the low-frequency-usage codons (rare codons) to the expression of the target gene in the host. In this study, we developed the method Presyncodon, available in a web version, to predict the gene code from a protein sequence, using built-in evolutionary information on a specific expression host. The synonymous codon-usage pattern of a peptide was studied from three genomic datasets (<i>Escherichia coli</i>, <i>Bacillus subtilis</i>, and <i>Saccharomyces cerevisiae</i>). Machine-learning models were constructed to predict a selection of synonymous codons (low- or high-frequency-usage codon) in a gene. This method could be easily and efficiently used to design new genes from protein sequences for optimal expression in three expression hosts (<i>E. coli</i>, <i>B. subtilis</i>, and <i>S. cerevisiae</i>). Presyncodon is free to academic and noncommercial users; accessible at http://www.mobioinfor.cn/presyncodon_www/index.html.