MSG: Seq [contigxxxx]: Terminator codon inside CDS!

[blancaverag]

I am trying to annotate metagenomic contigs with prokka. For that, I have joined bacteria, archaea and virus databases and run the prokka command. The computation runs fine until:

[05:38:20] Found 71830 CDS
[05:38:20] Connecting features back to sequences
[05:38:20] Not using genus-specific database. Try --usegenus to enable it.
[05:38:20] Annotating CDS, please be patient.
[05:38:20] Will use 8 CPUs for similarity searching.

--------------------- WARNING ---------------------
MSG: Seq [contig01018]: Terminator codon inside CDS!
...

What could be wrong?

[tseemann]

@blancaverag what version of Bioperl are you using? send me the whole .log file if possible.

[blancaverag]

Hi,

I have BioPerl 1.007000.
The problem arises when I add the --metagenome option. Without it, it just runs perfectly.
I am attaching the log file and the screen message, as the error does not appear in the log file.
PROKKA_11012016_log.txt
err_file.txt

[regelka]

Same problem here...

Prokka_log.txt

[mherold1]

I have noticed before that prodigal sometimes predicts CDS that contain stop codons, e.g. :

>PROKKA_00425 hypothetical protein
MTREKGYAYNRRFCASGAGRNNNEQFSKSQQWFGQDKQHSTLNHNFNKIFSIGSGCGQTE
FFFPA*HIALAVSGNCKPDNQNNQTLPKHEKFQKIHS*VFALSDCNSTVRSRVLEYLFWS
*CKSYFLSKSSCSYQFDLAVFITISTYSYRKETKRKS

[alneberg]

I've seen the same as @mherold1 when ran in the --metagenome mode.
I believe in our case it's because prodigal will disregard the -g flag when the metagenome mode is used.
Prodigal thus gives some genes from a different genetic code, and this information is lost since prokka only picks up the nucleotide sequence form prodigal.

Does that make sense?

[spock]

@alneberg , prokka does get only start/end/strand information from prodigal for each CDS.

[alneberg]

@spock, exactly, so then if prodigal have created a gene past a regular stop codon, there will be a stop codon in the amino acid sequence when prokka translates the nucleotide sequence with the original intended genetic code.

[arielamadio]

Could this be related with the issue I've added about selenocysteine?

[maureenbug]

I'm getting the same warning, and some of my samples have ~40% of the detected genes with internal stop codons... this seems quite high to me.

[willnotburn]

1.5 years late to this party. I understand what @alneberg is saying. OK, but then what? How does prokka resolve this inconsistency with prodigal? Should I still use the --metagenome mode? Using prokka 1.13

[tseemann]

Prokka was never intended to me used for metagenomes.

I did not know the Prodigal reported genes from different genetic codes!
That would be the source of the problem.

What exact version of prodigal are you using?
I think Homebrew and maybe Conda recently updated.

[willnotburn]

prokka log file says Prodigal version 2.6. Should it be a different one? I used the conda install just last week.

[tseemann]

@willnotburn Can you find the exact version?

% prodigal

-------------------------------------
PRODIGAL v2.6.3 [February, 2016]
Univ of Tenn / Oak Ridge National Lab
Doug Hyatt, Loren Hauser, et al.
-------------------------------------

Usage:  prodigal [-a trans_file] [-c] [-d nuc_file] [-f output_type]
                 [-g tr_table] [-h] [-i input_file] [-m] [-n] [-o output_file]
                 [-p mode] [-q] [-s start_file] [-t training_file] [-v]
         -g:  Specify a translation table to use (default 11).

[willnotburn]

@tseemann It is indeed PRODIGAL v2.6.3 [February, 2016]

[tseemann]

So it turns out that prodigal in metagenome mode will try translation tables 4 and 11 and see which does 'better'. So it might choose some contigs to be in table 4, but i don't know which ones are that. So when I go to translate proteins that were done with table 4, but I am using table 11 (from --gcode) then it might sometimes encounter stop codons.

Does prodigal tag which genetic code it uses for each prediction, in any of the 3 output modes?

[JinqunHuang]

@tseemann
Hi, Torsten,
WARNING massage "Terminator codon inside CDS" would be reported if I set the "--metagenome" in Prokka 1.13.

I try different parameter combination as below, the WARNING massages still appear.

"prokka --outdir AS100-2 --prefix AS100 --addmrna --metagenome --gcode 11 contigs.fasta.500"

or

"prokka --outdir AS100-3 --prefix AS100 --kingdom Bacteria --metagenome contigs.fasta.500"

Once I do not set the "--metagenome", the WARNING massages do not appear.
My datas are metagenomics datas, could you please have any advises ?

Prodigal V2.6.3
Prokka 1.13

Thanks a lot!
JinQun

[tseemann]

@JinqunHuang ok it seems prodigal ignores genetic code in metagenome mode, because it knows it is a mixture of different things. i can't tell what table prodigal used (?) so i don't know how to translate it correctly.

[cssulliv]

Hi,

New user here. I just installed Prokka via conda and I too am getting the "Terminator codon inside CDS" message but only for a subset of my MAGs when I include the --metagenome flag. My question is have other folks annotated their MAGs using a default prokka run and if so did it positively or negatively affect any downstream analysis? Any suggestions on this would be helpful :)

[tseemann]

The problem is that --metagenome invokes a special mode of prodigal (the gene predictor tool) which produces predictions which do NOT match --gcode 11. So you get stop codons (relative to 11) but they were predicted assuming some other code. Unfort prodigal does not record or indicate what code it used.

If you have nice contigs and they are all bacterial, don't use --metagenome.

[ScienceAdvances]

use --forece option will produce right *.faa file, but warnning message "Terminator codon inside CDS" still appear.
prokka Bin.01.fna --force --outdir ./ --prefix Bin.01 --metagenome --cpus 20 --kingdom Bacteria