I'm curious to find out which genes show an extremely biased AT content in human beings (i.e., within the same species, not compared to other ones). Any clue how I might find that out?
There is a very easy way to get all the human genes with their respective GC content (and therefore to deduce their AT content).
In Ensembl Biomart (http://www.ensembl.org/biomart/martview/), you have to select the human gene database, without any filters, and to select the %GC in the "Gene" attributes. This generates a file with all 60684 Genes know in Ensembl together with their GC content. You can add about any relevant info you may need about the genes, and further filter on other criteria, such as chromosome, type of genes (i.e. coding versus non-coding), and so on.
There is a very easy way to get all the human genes with their respective GC content (and therefore to deduce their AT content).
In Ensembl Biomart (http://www.ensembl.org/biomart/martview/), you have to select the human gene database, without any filters, and to select the %GC in the "Gene" attributes. This generates a file with all 60684 Genes know in Ensembl together with their GC content. You can add about any relevant info you may need about the genes, and further filter on other criteria, such as chromosome, type of genes (i.e. coding versus non-coding), and so on.
Considere when analyzing that there is a bias for the 5' region which is generally rich in GC, therefore shorter genes (genes with few exons and a short coding region) might have a higher GC content which is simply due to the 5' bias.