Codon optimisation¶

A protein can be spelled many ways in DNA, and a host does not use those spellings equally. One command writes a protein as the codons one host counts most often, then takes out every restriction site you name by swapping a codon for a synonym, so the protein itself is unchanged. It reports every codon it moved.

When to use it¶

• You are ordering a gene as synthetic DNA and want your host to read it well.
• A part must carry no BsaI, BbsI or other site, because the assembly cuts with it.
• You hold a coding sequence already and want to know which codons have to change.

A part written here goes straight into Golden Gate assembly, whose enzyme must not cut it, and barcode sets are kept clear of the same sites.

What you need¶

The protein itself, as one-letter amino acids typed on the command line rather than read from a file. A host with a shipped codon table: e-coli-k12, human or mouse. The protein below is GFP, read off a record that ships with the repo.

Run it¶

pixi run mbio codon-optimize --kind protein --host e-coli-k12 \
  --forbid BsaI --forbid NdeI --forbid NcoI --name GFP --out plan/ \
  MSKGEELFTGVVPILVELDGDVNGHKFSVSGEGEGDATYGKLTLKFICTTGKLPVPWPTLVTTFSYGVQCFSRYPDHMKRHDFFKSAMPEGYVQERTIFFKDDGNYKTRAEVKFEGDTLVNRIELKGIDFKEDGNILGHKLEYNYNSHNVYIMADKQKNGIKVNFKIRHNIEDGSVQLADHYQQNTPIGDGPVLLPDNHYLSTQSALSKDPNEKRDHMVLLEFVTAAGITHGMDELYK

Every option comes first, because the sequence is the last thing on the line. It prints the summary, then the two files it wrote:

GFP: 714 bp, 238 aa, host e-coli-k12, 3 codon change(s), free of BsaI, NdeI, NcoI
plan/coding-sequence.dna
plan/codon-changes.tsv

Three codons moved. Writing GFP with the codons E. coli counts most often spelled two NdeI sites and one NcoI site, and each went away when a histidine codon changed from CAT to CAC. BsaI was forbidden too and needed nothing, because those same codons spell no BsaI site here. Where no synonymous change can reach a site, the run stops rather than leave it in: a part carrying it would cut itself.

238 aa and 714 bp count no stop codon, because the protein handed over carries none. End the protein with * and a stop codon is written.

What it wrote¶

Both names link to what that run wrote, published here unedited. The same protein, host and enzymes always give the same bases, so the file is a rerun away rather than something to keep safe. To change the design, change the command and run it again.

How to read the sheet¶

codon-changes.tsv holds one row for each codon moved: where it sits in the protein, the codon before and after, the amino acid both of them spell, and the enzyme whose site the change cleared. There are three rows here, codons 76 and 216 for NdeI and codon 230 for NcoI.

All three are histidine, CAT to CAC. Histidine has two codons, so there was one synonym to move to. Read the amino acid column as the guarantee: the protein that came out is the protein that went in.

coding-sequence.dna is the sequence to send a vendor, and it opens in SnapGene like any other record. Check the name on it, which --name set, is the name you want on the order.

Before you order¶

• Check the host is the one the protein will be made in, not the one you clone in.
• Name every enzyme the part will meet, including one you use only in a later round.
• Check the protein you pasted is the whole protein, both ends included.
• Decide whether you need a stop codon, and end the protein with * if you do.
• Check it starts where translation should, because nothing adds a start codon for you.

Tips and troubleshooting¶

A site cannot be removed. The run stops and names the sequence, the site and the reason. No synonymous change reached it, so either change the protein over that site or cut with a different enzyme.

You already have a coding sequence. --kind dna checks what you hand it instead of writing it again. Its codons are left as they are, apart from any that has to move to clear a forbidden site.

You want the bases on the screen. Leave --out off, and the verb prints the summary, each codon it moved, and then the sequence, so it pipes into whatever comes next.

Where the numbers come from¶

A codon table here is a count of real coding sequences rather than a copy of a published compilation. E. coli K-12 MG1655 is counted over every complete coding sequence in its genome, fetched from NCBI. Human and mouse are counted over one canonical coding sequence per protein-coding gene, from GENCODE on the assemblies the lab holds.

So each table is the background usage of a whole genome. It is not a set of highly expressed genes, which is what a classical optimisation is built on. The genetic code that groups the codons is the standard one.

pixi run mbio codon-optimize --help lists the rest.

References¶

• NCBI E-utilities, spliced coding sequences for Escherichia coli str. K-12 substr. MG1655, GenBank U00096.3, read 14 September 2026.
• GENCODE v50 on GRCh38, one Ensembl canonical coding sequence per protein-coding gene, read 14 September 2026.
• GENCODE vM39 on GRCm39, the same for mouse, read 14 September 2026.
• The standard genetic code, NCBI translation table 1.