Ascertainment bias correction just for BIN and MULTISTATE?

[David Černý]

Hi everyone,

I am trying to run RAxML on a dataset that includes several nucleotide partitions, one partition consisting of binary characters, and another partition consisting of multistate characters. All of the characters included in the last two partitions are variable. I would like to analyze them as follows:

• nucleotides: GTR+gamma
  
• binary: Mk+gamma, Lewis' ascertainment bias correction
  
• multistate: Mk+gamma, Lewis' ascertainment bias correction

Is this actually possible? I tried specifying the corresponding models (DNA, ASC_BIN, and ASC_MULTI, respectively) in my partition file, and then running RAxML (v8.2.12) as follows:

./raxmlHPC-PTHREADS-AVX2 -f o -n combined -s infile.phy -m GTRGAMMA -K MK -q partitions.txt --asc-corr=lewis -p 12345 -N 50

However, this makes the program terminate with the following error:

IMPORTANT WARNING: Alignment column 16089 contains only undetermined values which will be treated as missing data

You are trying to infer per site likelihoods or ancestral states or do calculations with an ascertainment bias correction on an alignment containing 1 sites consisting only of undetermined characters. Please remove them first and then re-run RAxML!

The column in question belongs to one of the nucleotide partitions, so while I could solve the problem by removing it, the warning indicates that RAxML is trying to apply the specified ascertainment bias correction to all partitions, rather than just the partitions I told it to. Is there a way to restrict the correction to the binary and multistate partitions only – either in RAxML or in RAxML-NG?

Thank you,

--

David Černý

[Alexandros Stamatakis]

Dear David,

Could you maybe switch to RAxML-NG as standard RAxML is not supported
any more?

https://github.com/amkozlov/raxml-ng

Otherwise, if you insist on using standard RAxML, please let me know and
I will have a look what is happening there.

Alexis

On 13.02.21 21:21, David Černý wrote:
> Hi everyone,
>
> I am trying to run RAxML on a dataset that includes several nucleotide
> partitions, one partition consisting of binary characters, and another
> partition consisting of multistate characters. All of the characters
> included in the last two partitions are variable. I would like to
> analyze them as follows:
>

> * nucleotides: GTR+gamma
> * binary: Mk+gamma, Lewis' ascertainment bias correction
> * multistate: Mk+gamma, Lewis' ascertainment bias correction

>
> Is this actually possible? I tried specifying the corresponding models
> (DNA, ASC_BIN, and ASC_MULTI, respectively) in my partition file, and
> then running RAxML (v8.2.12) as follows:
>
> ./raxmlHPC-PTHREADS-AVX2 -f o -n combined -s infile.phy -m GTRGAMMA -K
> MK -q partitions.txt --asc-corr=lewis -p 12345 -N 50
>
> However, this makes the program terminate with the following error:
>
> IMPORTANT WARNING: Alignment column 16089 contains only undetermined
> values which will be treated as missing data
>
> You are trying to infer per site likelihoods or ancestral states or do
> calculations with an ascertainment bias correction on an alignment
> containing 1 sites consisting only of undetermined characters. Please
> remove them first and then re-run RAxML!
>
> The column in question belongs to one of the nucleotide partitions, so
> while I could solve the problem by removing it, the warning indicates
> that RAxML is trying to apply the specified ascertainment bias
> correction to all partitions, rather than just the partitions I told it
> to. Is there a way to restrict the correction to the binary and
> multistate partitions only – either in RAxML or in RAxML-NG?
>
> Thank you,
> --
> David Černý
>

> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/raxml/ef08c623-61b0-47b8-bf1f-5d5b5b899515n%40googlegroups.com
> <https://groups.google.com/d/msgid/raxml/ef08c623-61b0-47b8-bf1f-5d5b5b899515n%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology

www.exelixis-lab.org

[David Černý]

Dear Alexis,

Many thanks for your advice! Unlike standard RAxML, RAxML-NG has no problem with the combination of models and data types I described in my original post.

[Alexandros Stamatakis]

great :-)

alexis

> www.exelixis-lab.org <http://www.exelixis-lab.org>

>
>
>
> Thank you,
> --
> David Černý
>
> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/raxml/d9f59eb1-fec0-49b5-a577-685a5b2b82c5n%40googlegroups.com
> <https://groups.google.com/d/msgid/raxml/d9f59eb1-fec0-49b5-a577-685a5b2b82c5n%40googlegroups.com?utm_medium=email&utm_source=footer>.