Codon Usage Calculator — Frequency by Amino Acid

Because the genetic code is degenerate, 61 sense codons encode 20 amino acids, and most amino acids have two, four, or six synonyms. Relative Synonymous Codon Usage measures how far actual usage departs from the null expectation of equal use among synonyms. For a codon it is the observed count divided by the count expected if the total for that amino acid were split evenly across its synonymous family, so RSCU of 1.00 means no preference, values above 1 mean over-representation, and the ceiling equals the family size: 2 for a two-fold degenerate amino acid such as phenylalanine, 4 for a four-fold one such as glycine, 6 for leucine, serine, and arginine. Normalizing by family size is what makes RSCU comparable between genes that encode different proteins, since it removes the influence of amino acid composition entirely. Two companion statistics summarize the same sequence differently. The effective number of codons, Nc, expresses overall bias on an intuitive scale running from 20, when every amino acid uses exactly one codon, to 61, when all synonyms are used equally; most bacterial genes fall between about 30 and 55, with the lowest values in the most highly expressed genes. GC content at the third or wobble position is reported because that site is where synonymous change concentrates, and it tracks genome-wide compositional pressure: Streptomyces coelicolor sits above 90 percent GC3 while Plasmodium falciparum falls near 10 percent. Interpreting any of these numbers from a short sequence is unreliable, since RSCU on a handful of codons is dominated by sampling noise, and Nc requires several counts per amino acid family to be stable. Bias also reflects two entangled causes, mutational pressure and translational selection, which these descriptive statistics cannot separate on their own.

Systematic study of synonymous codon choice began once enough sequences existed to count. Through the early 1980s Toshimichi Ikemura demonstrated in Escherichia coli and Saccharomyces cerevisiae that preferred codons match abundant tRNA species, giving codon bias a translational explanation rather than a purely mutational one. RSCU was introduced by Paul Sharp, Therese Tuohy, and Krzysztof Mosurski in Nucleic Acids Research in 1986, in a survey of codon usage across organisms, precisely to allow comparison of genes with different amino acid content. The effective number of codons was proposed by Frank Wright in the journal Gene in 1990, adapting the effective number of alleles from population genetics so that bias could be quantified without reference to any set of highly expressed genes; the Nc against GC3 plot he introduced remains a standard diagnostic for distinguishing mutational from selective bias. Reference data for these calculations were centralized in the Codon Usage Database at the Kazusa DNA Research Institute, compiled from GenBank by Yasukazu Nakamura and colleagues and described in a 2000 Nucleic Acids Research report, which tabulated usage for thousands of organisms. The field turned applied in the 2000s: as commercial gene synthesis costs fell, whole coding sequences were routinely redesigned to match host preferences, and studies that followed complicated the simple picture by showing that maximizing preferred codon use can deplete specific tRNA pools or eliminate deliberate translational pauses, so current practice balances codon choice against mRNA folding and GC content rather than optimizing usage alone.