Codon Usage Database Tabulated from GenBank data

Question

[Bohdan Khomtchouk]

Can anyone explain how the codon usage table database for Homo sapiens is composed of 93,487 CDS's (source: http://www.kazusa.or.jp/codon/cgi-bin/showcodon.cgi?species=9606), yet there are only 32,109 instances of the codon UAG? This would mean that certain CDS's did not contain the UAG codon, which does not make sense. There are other "counterintuitive" codons like this in the database table...

[5heikki]

In the majority of the genetic codes, UAG is a stop codon so generally e.g. human CDS would at most have 1 UAG codon.

[Bohdan Khomtchouk]

Even if you sum the absolute counts of all the stop codons together, you get more stop codons than CDS sequences, which is counter-intuitive (as per the definition of CDS). Thoughts?

[5heikki]

UGA codes tryptophan in human mitochondria. More than that, stop codons do not necessarily always lead to stop of translation but can e.g. induce a frameshift on rare occasion. Also, that CDS count seems quite high and likely includes pseudo genes.

Answer 1

[Antonio R. Franco]

Differential splicing? Most human genes do it.

[Bohdan Khomtchouk]

Differential splicing seems relevant in this DNA-level codon usage analysis... because how else would they calculate codon usage other than to evaluate the RNA output and count how many RNA's of each kind are produced (i.e., how many codons are actively transcribed from each CDS). This is the only way to rationalize the "counter-intuitive" numbers. This would make the first two sentences of the original paper abstract make sense: http://www.ncbi.nlm.nih.gov/pubmed/10592250

Thoughts?