Alternative non-templated codon usage?

[JamieHeather]

Currently stitchr uses Kasuza frequency data to pick codons for residues that are not wholly germline-encodable when providing AA CDRs. This basically picks the most common nucleotide triplet used from the relevant species' exome.

I'm wondering whether it might be useful in certain circumstances to instead draw on codon frequency usage taken from non-templated sections of CDR3s of the corresponding species/locus. This might theoretically deliver NT sequences that are more representative of rearranged TCRs of the specific locus, which are not necessarily the same as the wider exome.