Dear Jeong,
> Thank you very much for your detailed explanations and summaries! Though
> I am still unsure what I can do to improve the results in relation to
> AIC for mPTP, I now understand that PTP and single-rate mPTP are
> essentially the same, minus the different heuristic algorithm. I suppose
> I should stick with the single rate for now.
We believe that the multiple rate version is the better model.
> Would you be so kind to look at my other question regarding PTP (and
> mPTP)? Why is it that removing distantly related groups improves the
> results? For my dataset, PTP only delimited the closely related species
> accurately when I removed the outgroups and other distantly related
> ingroup species.
Well, as we don't know the absolute truth, I would state that removing
the outgroups yields different results, not necessarily improve results.
Beyond this, both PTP and mPTP are essentially conducting a
classification on the branch lengths of a phylogenetic tree into inter
versus intra-species branches, hence fiddling with the branch lengths or
even the tree itself (removing the outgroup) will change the branch
length distribution that mPTP operates on and hence naturally yield
different results. I nontheless, do believe that removing an outgroup
yet keeping the root it yielded is a reasonable thing to do such as to
not distort mpTP too much.
Hope this helps,
Alexis
>
> Regards,
> Jeong
>
>
> On Tuesday, March 29, 2022 at 2:51:36 PM UTC-4
sarah.l...@h-its.org wrote:
>
> Hi Jeong!
>
>
> Back then we implemented two target functions for mPTP, using either
> single rate or multiple rates:
>
>
> In mPTP single, we compute the loglikelihood the same way as in the
> PTP program, using a single rate.
>
> In mPTP multi, we compute the loglikelihood using multiple rates.
>
>
> From the command line documentation
> (
https://github.com/Pas-Kapli/mptp/blob/master/src/mptp.c#L390
> <
https://github.com/Pas-Kapli/mptp/blob/master/src/mptp.c#L390>):
> <
https://github.com/Pas-Kapli/mptp/blob/master/src/dp.c> and
> <
https://github.com/Pas-Kapli/mptp/blob/master/src/dp.c#L237>
> <
https://en.wikipedia.org/wiki/Akaike_information_criterion#Modification_for_small_sample_size>)
> <
https://github.com/Pas-Kapli/mptp/blob/master/src/dp.c#L194>). With
> this criterion, a more complex model (i.e., more species) is only
> accepted if it increases the target function's value enough compared
> to a less complex model (i.e., less species). This is why you
> observed over-lumping on your data when using mPTP with this model.
>
>
> To summarize:
>
> - Both PTP and mPTP-single optimize exactly the same target
> function. The only difference lies in the heuristic algorithm that
> is used to maximize the loglikelihood under a single rate.
>
> - mPTP-multi does model comparison via AIC, which can lead to
> over-lumping in some cases.
>
>
> I hope this helps! Sorry that I did not remember these details. The
> code is nearly 7 years old, and I'm not working in Academia anymore.
>
>
> Best regards,
>
> Sarah
>
>
> ------------------------------------------------------------------------
> *From:*
ptp-species-...@googlegroups.com
> *Sent:* Tuesday, March 29, 2022 8:48 AM
>
> *To:* PTP and GMYC species delimitation
> *Subject:* Re: PTP in bPTP web server vs mPTP web server
> ------------------------------------------------------------------------
> *From:*
ptp-species-...@googlegroups.com
> *Sent:* Tuesday, March 29, 2022 3:05 AM
> *To:* PTP and GMYC species delimitation
> *Subject:* PTP in bPTP web server vs mPTP web server
> Dear community,
>
> I've recently delved into PTP, and tried it on two webservers
> (
https://mptp.h-its.org/ <
https://mptp.h-its.org/> and
>
https://species.h-its.org/ <
https://species.h-its.org/>).
>
> Are the single rate PTP in mptp server and PTP in the other
> server different from one another?
>
> In mptp website, I tried single-rate, which resulted in
> partitions close to morphology and ABGD. (55 partitions)
>
> In PTP website however, both maximum likelihood and bayesian
> solution resulted in unrealistic oversplitting of species. (over
> 8- partitions)
>
> I would appreciate your helps.
>
> --
> You received this message because you are subscribed to the
> Google Groups "PTP and GMYC species delimitation" group.
> To unsubscribe from this group and stop receiving emails from
> it, send an email to
ptp-species-delimi...@googlegroups.com.
> To view this discussion on the web visit
>
https://groups.google.com/d/msgid/ptp-species-delimitation/4fbdc802-cb8d-4fb0-871f-9df5b72575fen%40googlegroups.com
> <
https://groups.google.com/d/msgid/ptp-species-delimitation/4fbdc802-cb8d-4fb0-871f-9df5b72575fen%40googlegroups.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to the Google
> Groups "PTP and GMYC species delimitation" group.
> To unsubscribe from this group and stop receiving emails from it,
> send an email to
ptp-species-delimi...@googlegroups.com.
> To view this discussion on the web visit
>
https://groups.google.com/d/msgid/ptp-species-delimitation/7d357ba6-3c7f-4909-a943-4d95ac334337n%40googlegroups.com
> <
https://groups.google.com/d/msgid/ptp-species-delimitation/7d357ba6-3c7f-4909-a943-4d95ac334337n%40googlegroups.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to the Google
> Groups "PTP and GMYC species delimitation" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to
ptp-species-delimi...@googlegroups.com
> <mailto:
ptp-species-delimi...@googlegroups.com>.
> To view this discussion on the web visit
>
https://groups.google.com/d/msgid/ptp-species-delimitation/1b9ea4a2-26f6-4750-a25c-3af93ddc3a11n%40googlegroups.com
> <
https://groups.google.com/d/msgid/ptp-species-delimitation/1b9ea4a2-26f6-4750-a25c-3af93ddc3a11n%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
Affiliated Scientist, Evolutionary Genetics and Paleogenomics (EGP) lab,
Institute of Molecular Biology and Biotechnology, Foundation for
Research and Technology Hellas
www.exelixis-lab.org