Model Partition in RAxML-NG Tree Calculation

[EB Wong]

Hi,

I am creating model partition files for all my concatenated alignments, but wondering about the way RAxML-NG treated the partitions.
Say if there are 3 genes in the concatenated alignment, each with 100 bases, the first two are to be specified with model A and the third one with model B.

I have been splitting the models for each of the partition (as in ModelFile1).

Wondering will there be any difference in the tree calculation if the consecutively same models being combined as single model partition (as in ModelFile2)?

Does RAxML-NG apply the models independently in each site for the tree calculation, that it does not impact much even when the partitions being combined?

Sample partitioned model files for RAxML-NG

ModelFile1:
LG, gene1=1-100
LG, gene2=101-200
WAG, gene3=201-300

ModelFile 2:
LG, gene1gene2=1-200

WAG, gene3=201-300

Thank you,

Bhei

[Alexey Kozlov]

Hi,

combining multiple genes with the same substitution model into a single partition would result in
joint estimation of free model parameters (e.g., alpha parameter for +G). So in general, it will
make a difference for likelihood computation, except if there are no free parameters to estimate.

Please see here for details:

https://github.com/amkozlov/raxml-ng/wiki/Tutorial#partitioned-analysis

https://github.com/amkozlov/raxml-ng/wiki/Input-data#multiple-models

On 12.09.22 09:47, EB Wong wrote:
> Hi,
>
> I am creating model partition files for all my concatenated alignments, but wondering about the way
> RAxML-NG treated the partitions.
> Say if there are 3 genes in the concatenated alignment, each with 100 bases, the first two are to be
> specified with model A and the third one with model B.
>
> I have been splitting the models for each of the partition (as in ModelFile1).
> Wondering will there be any difference in the tree calculation if the consecutively same models
> being combined as single model partition (as in ModelFile2)?
>
> Does RAxML-NG apply the models independently in each site for the tree calculation, that it does not
> impact much even when the partitions being combined?
>

> _Sample partitioned model files for RAxML-NG_

> ModelFile1:
> LG, gene1=1-100
> LG, gene2=101-200
> WAG, gene3=201-300
>
> ModelFile 2:
> LG, gene1gene2=1-200
> WAG, gene3=201-300
>
> Thank you,
> Bhei
>

> --
> 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/9e7e22a5-494c-405c-ac43-94710d6e36d8n%40googlegroups.com
> <https://groups.google.com/d/msgid/raxml/9e7e22a5-494c-405c-ac43-94710d6e36d8n%40googlegroups.com?utm_medium=email&utm_source=footer>.