Stop codon

[Tristan Lefebure]

Hi there,
I am analyzing a gene that gets non-functional in some taxa and therefore accumulates many STOP codons in the middle of the alignment. These STOP codon are very informative, and I want to keep them in the analyses, and just consider them as non-synonymous substitutions. Using the latest testNH code, this does not seem possible anymore:

- in bppml, if I set input.sequence.remove_stop_codons = no, I get a crash like this:

Number of distinct sites...............: 260
WARNING!!! !!! Warning!!! Initial likelihood is zero.
WARNING!!! !!! This may be due to branch length == 0.
WARNING!!! !!! All null branch lengths will be set to 0.000001.
Initial log likelihood.................: -inf
ERROR!!! !!! Unexpected initial likelihood == 0.
Stop Codon at site 17 in sequence PguiUba1
Stop Codon at site 22 in sequence PdiaCli1
Stop Codon at site 22 in sequence ParaBur1
Stop Codon at site 35 in sequence ProuMes1
...

- with the same option in mapnh, I am getting a non-informative star-tree clustering and the output looks like:
----
Remove Stop Codons.....................: no
Number of sequences....................: 60
Number of sites........................: 260
Tree file..............................: op2_1motu_proas.ml_h.dnd
Number of leaves.......................: 60
Heterogeneous model....................: no
Saturation threshold used..............: 3
WARNING!!! On branch 0, counts could not be computed.
WARNING!!! On branch 1, counts could not be computed.
WARNING!!! On branch 2, counts could not be computed.
WARNING!!! On branch 3, counts could not be computed.
WARNING!!! On branch 4, counts could not be computed.
WARNING!!! On branch 5, counts could not be computed.
WARNING!!! On branch 6, counts could not be computed.
WARNING!!! On branch 7, counts could not be computed.
WARNING!!! On branch 8, counts could not be computed.
WARNING!!! On branch 9, counts could not be computed.
WARNING!!! On branch 10, counts could not be computed.
WARNING!!! On branch 11, counts could not be computed.
WARNING!!! On branch 12, counts could not be computed.
WARNING!!! On branch 13, counts could not be computed.
WARNING!!! On branch 14, counts could not be computed.
WARNING!!! On branch 15, counts could not be computed.
WARNING!!! On branch 16, counts could not be computed.
WARNING!!! On branch 17, counts could not be computed.
WARNING!!! On branch 18, counts could not be computed.
WARNING!!! On branch 19, counts could not be computed.
...

How can I revert to the original behaviour and include the stop codons in the analysis (and also the sites with stop codons)?

Thanks!
--
Tristan

[Julien Yann Dutheil]

Hi Tristan,

I think this has never worked: codon model do not allow the occurrence
of stop codons in the sequences (but Laurent can answer this better
than me).
I am afraid you have to replace them by XXX if you still want to use the site.

Cheers,

Julien.

> --
>
>
>



--
Julien Y. Dutheil, Ph-D
0 (+49) 6421 178 530

§ Max Planck Institute for Terrestrial Microbiology
Department of Organismic Interactions
Marburg -- GERMANY

§ Intitute of Evolutionary Sciences - Montpellier
University of Montpellier 2 -- FRANCE

[Tristan Lefebure]

Hi,
Yes, I was wrong, my bad, It's indeed what I was already doing: masking the stops before using testNH.

If one would make a new fake genetic code where any STOP codons would instead code for a new fake amino-acid (or may be an amino acid that is not encoded by other codons that can be reached with one silent substitution), and define it in say /bpp-seq/src/Bpp/Seq/GeneticCode/NoStopGeneticCode.bpp, would it be possible to use it directly within testNH?

(I know, I live in a very naive world)

Thanks!
--
tristan

[Laurent Guéguen]

Hi Tristan,

in the usual codon models, indeed stop codons are not considered,
perhaps because it is conceptually difficult to link it with
synonym/non-synonym definition.

Yes indeed, it is very easy to define an adhoc "extended" genetic
code,
but in my point of view it is not correct to define an artificial
'stop' amino-acid
that would behave as a usual one. So I could develop a model with an
additional coding vs non-coding triplet substitution rate. If you need
it,
it is very easy for me to do this, we can talk about it.

Cheers,
Laurent