[Feature Request] Allow +ASC for mixture models

[HS6986]

Dear IQ-TREE Developers,

The ascertainment bias correction parameter +ASC is currently not applicable to mixture models, and attempts to apply it result in the message Mixture model +ASC is not supported yet. Contact author if needed. However, I think that it would be valuable to support it.

It would be useful when applying mixture models, for example, to binary microsynteny data without invariant characters (e.g., Vasilikopoulos et al., 2024) and neomorphic (present/absent; see Sereno, 2007) binary morphological characters (e.g., Juravel et al., 2023). I am sure there are other situations where this would be useful.

Although of course we would have to give this proposal a lower priority than the others, I would support promoting it.

[bqminh]

Is this now addressed in your MixtureFinder pull request #11 ?

[HS6986]

Dear bqminh,

No, it isn't. Honestly, I don't understand how an ascertainment bias correction in IQ-TREE works internally at all, and I don't think I can handle implementing ascertainment bias corrections in mixture models.

By the way, I've perhaps succeeded in implementing a new MixtureFinder-like algorithm for selecting the best-fitting mixture models for morphological data in IQ-TREE, which has not yet been validated in publications. I'll post it later. Ascertainment bias corrections in mixture models would of course also be useful (or possibly even required) in this framework. I would really appreciate it if someone could implement them when they are available.