Support start codons in translation tables

[fungs]

Hi, thanks again for this wonderful software, with every new release it replace more programs in my pipelines. Now for this simple feature request.

Translation tables come with alternative start codons: https://www.ncbi.nlm.nih.gov/Taxonomy/Utils/wprintgc.cgi?mode=c#SG11

Would it be possible to treat the lookup for the start codon (the first codon of the input sequence) differently to map the correct amino acid? When I map sequences with correct frame and offset to protein space, it happens that some use alternative genetic codes (e.g. plastid-encoded onces) which leads to a wrong translation of the first codon. That could be an additional command line switch and would only makes sense for complete gene sequences.

Best,
Johannes

[shenwei356]

Can you give ome example with sequence?

[shenwei356]

One codon in a given translation table (-T/--transl-table) is only translated to one amino acid, no matter whether it's start codon or not.

[fungs]

I know, this should be marked as a feature request.

Translation tables have alternative amino acids (basically M) for start codons. The standard table doesn't. It's documented under the link I posted.

Best,
Johannes

[fungs]

Sorry, I didn't read your example request, will post you some today!

[fungs]

As an example, see locus ERS432292_01433 in CWZP01000026.

Protein sequence:
MDSPAEYKIFLNIILAPVDDGLRFTYSGDNFFYREISMISLIAALAVDRVIGMENAMPWNLPADLAWFKRNTLNKPVIMGRHTWESIGRPLPGRKNIILSSQPGTDDRVTWVKSVDEAIAACGDVPEIMVIGGGRVYEQFLPKAQKLYLTHIDAEVEGDTHFPDYEPDDWESVFSEFHDADAQNSHSYCFEILERRYFCIEFTASAGCDAGRVLSGLPISGCV

Nucleotide sequence:
GTGGACTCGCCAGCAGAATATAAAATTTTCCTCAACATCATCCTCGCACCAGTCGACGACGGTTTACGCTTTACGTATAGTGGCGACAATTTTTTTTATCGGGAAATCTCAATGATCAGTCTGATTGCGGCGTTAGCGGTAGATCGCGTTATCGGCATGGAAAACGCCATGCCGTGGAACCTGCCTGCCGATCTCGCCTGGTTTAAACGCAACACCTTAAATAAACCCGTGATTATGGGCCGCCATACTTGGGAATCAATCGGTCGTCCGTTGCCAGGACGCAAAAATATTATCCTCAGCAGTCAACCGGGTACGGACGATCGCGTAACGTGGGTGAAGTCGGTGGATGAAGCCATCGCGGCGTGTGGTGACGTACCAGAAATCATGGTGATTGGCGGCGGTCGCGTTTATGAGCAGTTCCTGCCAAAAGCGCAGAAACTGTATCTGACGCATATCGACGCAGAAGTGGAAGGCGACACCCATTTCCCGGATTACGAGCCGGATGACTGGGAATCGGTATTCAGCGAATTCCACGATGCTGATGCGCAGAACTCTCACAGCTATTGCTTTGAGATTCTGGAGCGGCGGTACTTTTGTATAGAATTTACGGCTAGTGCCGGATGCGACGCCGGTCGCGTCTTATCCGGCCTTCCTATATCAGGCTGTGTT

Translation with table 11, as given in the genome annotation, with seqkit gives:
VDSPAEYKIFLNIILAPVDDGLRFTYSGDNFFYREISMISLIAALAVDRVIGMENAMPWNLPADLAWFKRNTLNKPVIMGRHTWESIGRPLPGRKNIILSSQPGTDDRVTWVKSVDEAIAACGDVPEIMVIGGGRVYEQFLPKAQKLYLTHIDAEVEGDTHFPDYEPDDWESVFSEFHDADAQNSHSYCFEILERRYFCIEFTASAGCDAGRVLSGLPISGCV

Note that the translation table specifies GUG as an alternative start codon.

[shenwei356]

I get it, we should mark start codon's product as 'M' when it acts as start codon.

• seqkit_windows_amd64.exe.tar.gz
• seqkit_darwin_amd64.tar.gz
• seqkit_linux_amd64.tar.gz

$ seqkit seq t.fa 
>seq
GTGGACTCGCCAGCAGAATATAAAATTTTCCTCAACATCATCCTCGCACCAGTCGACGAC
GGTTTACGCTTTACGTATAGTGGCGACAATTTTTTTTATCGGGAAATCTCAATGATCAGT
CTGATTGCGGCGTTAGCGGTAGATCGCGTTATCGGCATGGAAAACGCCATGCCGTGGAAC
CTGCCTGCCGATCTCGCCTGGTTTAAACGCAACACCTTAAATAAACCCGTGATTATGGGC
CGCCATACTTGGGAATCAATCGGTCGTCCGTTGCCAGGACGCAAAAATATTATCCTCAGC
AGTCAACCGGGTACGGACGATCGCGTAACGTGGGTGAAGTCGGTGGATGAAGCCATCGCG
GCGTGTGGTGACGTACCAGAAATCATGGTGATTGGCGGCGGTCGCGTTTATGAGCAGTTC
CTGCCAAAAGCGCAGAAACTGTATCTGACGCATATCGACGCAGAAGTGGAAGGCGACACC
CATTTCCCGGATTACGAGCCGGATGACTGGGAATCGGTATTCAGCGAATTCCACGATGCT
GATGCGCAGAACTCTCACAGCTATTGCTTTGAGATTCTGGAGCGGCGGTACTTTTGTATA
GAATTTACGGCTAGTGCCGGATGCGACGCCGGTCGCGTCTTATCCGGCCTTCCTATATCA
GGCTGTGTT
>seq2
GTGGACTCGCCAGCAtaaGTGGACTCGCCAGCA
>seq2
GTGGACTCGCCAGCAtaattgGTGGACTCGCCAGCA

$ cat t.fa | seqkit translate -T 11
>seq
MDSPAEYKIFLNIILAPVDDGLRFTYSGDNFFYREISMISLIAALAVDRVIGMENAMPWN
LPADLAWFKRNTLNKPVIMGRHTWESIGRPLPGRKNIILSSQPGTDDRVTWVKSVDEAIA
ACGDVPEIMVIGGGRVYEQFLPKAQKLYLTHIDAEVEGDTHFPDYEPDDWESVFSEFHDA
DAQNSHSYCFEILERRYFCIEFTASAGCDAGRVLSGLPISGCV
>seq2
MDSPA*MDSPA
>seq2
MDSPA*MVDSPA

[fungs]

I get it, we should mark start codon's product as 'M' when it acts as start codon.

I think to do this correctly, you need to assume (1) that genes are complete so that the first base it actually a start codon, and (2) check whether the first codon is a valid start codon for the specified translation table.

I don't know where and how you keep the translation tables but this information needs to be added I suppose. For instance, table 11 has the following extra start codons: TTG, CTG, ATT, ATC, ATA, GTG in addition to ATG.

BTW: there is a similar issue with alternative stop codons for some of the more more exotic translation tables

[shenwei356]

The tables are here, start codons are flagged. All data come from https://www.ncbi.nlm.nih.gov/Taxonomy/taxonomyhome.html/index.cgi?chapter=cgencodes.

(1) I assume the gene are complete; (2) done.

[shenwei356]

@fungs any concern? can I close this issue?

[fungs]

@shenwei356 that is great, I will compile and test the current master next week and report if things are working well. Thx!

[fungs]

Hi @shenwei356,

I tested the current master flagged as v10.0.1 (note: it says "New version available: seqkit v0.10.0") on a larger set of protein coding nucleotide sequences. The deviation of translated versus deposited protein sequences is reduced and the program seems to work very well. I do however have some suggestions for safer handling.

1. Currently, the start codon auto-detection triggers automatically without the possibility to turn if off. I'd advice to not enable this by default and instead add a new command line parameter. Otherwise the feature might snap in unintendedly when translating partial gene sequences that don't start at the beginning.

2. I saw that the auto-detection triggers both at the beginning of a sequence and after a stop codon. I had, however, sequences which seem to contain a faulty stop codon (either due to sequencing error or due to an undocumented/novel genetic code). This then may alter the consecutive amino acid. One example is a variant of gene AJ131405 I have, where the sequence TTATAGATC is annotated as ...QI... but becomes ...*M... with translation table 11.

So basically it works very well but I think it needs a little better control.

Best,
Johannes

[shenwei356]

1. new option -M, --init-codon-as-M translate initial codon at beginning to 'M'
2. hard to solve.

[fungs]

Great work! I guess 2 can be neglected, the behavior is documented here now. It's a very special case which is only of importance when working with novel and erroneous data.

Best,
Johannes