Some questions about models and ancestral state reconstruction

[Vincent Lynch]

Hi All,

I'm using IQ-Tree to do ASR on a binary encoded morphological dataset and have some questions.

1)  ModelFinder indicates the best model is GTR2, but I cannot find an explanation of what this model is, is it described anywhere?.

2) My dataset includes missing data coded as "?", rather than skipping columns with ? IQ-Tree does the ASR on these sites. Is that normal? 

3) The IQ-Tree ASR does get] it "right" most of the time when a column includes ? sites. Can it also predict whether a ? in an extant species is likely to be either 0 or 1? If so how can one access this data?

Thanks!

Vinny

[Minh Bui]

Hi Vinny,

On 7 Oct 2019, at 2:19 am, Vincent Lynch <vincent...@gmail.com> wrote:

Hi All,

I'm using IQ-Tree to do ASR on a binary encoded morphological dataset and have some questions.

1)  ModelFinder indicates the best model is GTR2, but I cannot find an explanation of what this model is, is it described anywhere?.

This is the general time reversible model for binary data (that’s why the suffix “2”). It’s the same merit as GTR for DNA, just adapted for 2 character states.

2) My dataset includes missing data coded as "?", rather than skipping columns with ? IQ-Tree does the ASR on these sites. Is that normal? 

Yes that’s normal. While some taxa has “?”, other taxa having “proper” states will help to predict the ancestral states. For example, if a site is “???0????”, then IQ-TREE can still infer the ancestral states (likely 0 in this example).

3) The IQ-Tree ASR does get] it "right" most of the time when a column includes ? sites. Can it also predict whether a ? in an extant species is likely to be either 0 or 1? If so how can one access this data?

Good question. Right now there is no such option, as ASR is meant for (extinct) internal nodes. But why do you need such a feature? If there is a good reason why it’d be useful, then we will consider to implement that.

Cheers

Minh

Thanks!

Vinny

--
You received this message because you are subscribed to the Google Groups "IQ-TREE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iqtree+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/iqtree/b7ffdb50-902f-43d2-8c21-065e645009de%40googlegroups.com.

[Vincent Lynch]

Hi Minh,

Thanks for the quick reply!

This is the general time reversible model for binary data (that’s why the suffix “2”). It’s the same merit as GTR for DNA, just adapted for 2 character states.

 

Good question. Right now there is no such option, as ASR is meant for (extinct) internal nodes. But why do you need such a feature? If there is a good reason why it’d be useful, then we will consider to implement that.

Our dataset is a matrix of gene expression coded as 0 for "not expressed", 1 for "expressed", and ? for "missing data". In this case the data is missing because those genes likely have not been annotated in particular genomes; we'd like to be able to impute gene expression info using the ancestral reconstructions because while those data are missing they can still provide useful information (for example, if the BPP of the inference is >0.8 or something we include the imputed value in downstream analyses, otherwise we consider it a poorly imputed inference and exclude it). I think such a phylogenetic imputation feature would also just be generally useful. 

If it is implemented we're happy to include you and others as co-authors on manuscripts in which we use the data. 

Best,

Vinny

 

Cheers

Minh

Thanks!

Vinny

--
You received this message because you are subscribed to the Google Groups "IQ-TREE" group.

To unsubscribe from this group and stop receiving emails from it, send an email to iqt...@googlegroups.com.