Application to calculate consensus tree of RRHS from SNP data

[dandaman]

Hi,

this is not a real issue, but rather a question on the applicability of ASTRAL-II:

I wonder if it is sensible to apply the coalescence approach implemented in ASTRAL to generate a consensus topology for replicate maximum likelhood trees generated with the RRHS (repeated random haplotype samples) following a procedure similar to https://academic.oup.com/mbe/article/31/4/817/1100394?

The data is genome-wide LD-filtered and represents 1000 unique random samples of the same sites (variability emerging from heterozygous sites).

Any comment or suggestion is greatly appreciated.

Best,
Daniel

[dandaman]

Ah, forgot to add that the resulting tree topology and branch lengths in my data appear to make (biological) sense...

[smirarab]

Hi Daniel, I hadn't seen the RRHS approach before and I am just starting to read it to understand how it works. So the short answer is I am not sure. Can you tell me how you are using RRHS with ASTRAL though? Are you using it to combine trees from multiple random selections (e.g., instead of the majority rule consensus in the original paper)? Or are you somehow making gene trees? In other words, is each input tree to ASTRAL estimated from data coming from the whole genome or is it from an individual locus? Thanks Siavash

…

On Tue, Sep 25, 2018 at 12:17 PM Daniel Lang ***@***.***> wrote: Ah, forgot to add that the resulting tree topology and branch lengths in my data appear to make (biological) sense... — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#38 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAybuGlKoGMPRQ34QvyAWMYTVw027xgDks5ueoFjgaJpZM4W5ToZ> .

-- Siavash Mirarab

[dandaman]

Hi Siavash, thank you for your response. I'm using astral to generate a consensus tree. Does this makes sense? As I wrote, topologies look great, although some branch lengths appear longer than in other approaches I applied to the same data... The data set liekly contains a lot of reticulation introgression events. Best, Daniel Am 26. September 2018 20:34:07 MESZ schrieb Siavash Mirarab <notifications@github.com>:

…

Hi Daniel, I hadn't seen the RRHS approach before and I am just starting to read it to understand how it works. So the short answer is I am not sure. Can you tell me how you are using RRHS with ASTRAL though? Are you using it to combine trees from multiple random selections (e.g., instead of the majority rule consensus in the original paper)? Or are you somehow making gene trees? In other words, is each input tree to ASTRAL estimated from data coming from the whole genome or is it from an individual locus? Thanks Siavash On Tue, Sep 25, 2018 at 12:17 PM Daniel Lang ***@***.***> wrote: > Ah, forgot to add that the resulting tree topology and branch lengths in > my data appear to make (biological) sense... > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#38 (comment)>, or mute > the thread > <https://github.com/notifications/unsubscribe-auth/AAybuGlKoGMPRQ34QvyAWMYTVw027xgDks5ueoFjgaJpZM4W5ToZ> > . > -- Siavash Mirarab -- You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: #38 (comment)

-- Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.

[smirarab]

Daniel, Several points. 1. ASTRAL is usually used to combine gene trees with different true histories, as you probably know. 2. However, it probably is ok to use ASTRAL just as a way to summarize a bunch of alternative trees. Your input trees don't reflect gene tree discordance or anything like that. What you need, essentially, is just a supertree method to find a consensus among input trees, and ASTRAL can act as a supertree. But make sure not to interpret the results as a coalescent-based analysis (which it will not be). So the ASTRAL tree you will get with this input will not be taking into account ILS. (at least not in any obvious way). 3. Related to point 2, the way ASTRAL computes branch lengths is *very* dependent on ILS assumptions. So, you should probably ignore branch lengths. Ditto for localPP. You don't want to use them. 4. If you run ASTRAL with -t 1 it gives you quartet scores for each branch. The quartet score gives you a measure of the amount of discordance with each branch among your input trees. This is likely going to be useful for purposes. Thanks Siavash On Wed, Sep 26, 2018 at 12:18 PM Daniel Lang <notifications@github.com> wrote:

…

Hi Siavash, thank you for your response. I'm using astral to generate a consensus tree. Does this makes sense? As I wrote, topologies look great, although some branch lengths appear longer than in other approaches I applied to the same data... The data set liekly contains a lot of reticulation introgression events. Best, Daniel Am 26. September 2018 20:34:07 MESZ schrieb Siavash Mirarab < ***@***.***>: >Hi Daniel, > >I hadn't seen the RRHS approach before and I am just starting to read >it to >understand how it works. So the short answer is I am not sure. > >Can you tell me how you are using RRHS with ASTRAL though? Are you >using it >to combine trees from multiple random selections (e.g., instead of the >majority rule consensus in the original paper)? >Or are you somehow making gene trees? >In other words, is each input tree to ASTRAL estimated from data coming >from the whole genome or is it from an individual locus? > >Thanks >Siavash > >On Tue, Sep 25, 2018 at 12:17 PM Daniel Lang ***@***.***> >wrote: > >> Ah, forgot to add that the resulting tree topology and branch lengths >in >> my data appear to make (biological) sense... >> >> — >> You are receiving this because you are subscribed to this thread. >> Reply to this email directly, view it on GitHub >> ><#38 (comment)>, >or mute >> the thread >> >< https://github.com/notifications/unsubscribe-auth/AAybuGlKoGMPRQ34QvyAWMYTVw027xgDks5ueoFjgaJpZM4W5ToZ > >> . >> > > >-- >Siavash Mirarab > > >-- >You are receiving this because you authored the thread. >Reply to this email directly or view it on GitHub: >#38 (comment) -- Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#38 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAybuItGBRlXK0DEAWXX1EHK2K7gYGy4ks5ue9LwgaJpZM4W5ToZ> .

-- Siavash Mirarab

[dandaman]

Thanks so much for these helpful explanations. I'll use the quartet scores and no branch lengths then. Best, Daniel Am 26. September 2018 21:58:57 MESZ schrieb Siavash Mirarab <notifications@github.com>:

…

Daniel, Several points. 1. ASTRAL is usually used to combine gene trees with different true histories, as you probably know. 2. However, it probably is ok to use ASTRAL just as a way to summarize a bunch of alternative trees. Your input trees don't reflect gene tree discordance or anything like that. What you need, essentially, is just a supertree method to find a consensus among input trees, and ASTRAL can act as a supertree. But make sure not to interpret the results as a coalescent-based analysis (which it will not be). So the ASTRAL tree you will get with this input will not be taking into account ILS. (at least not in any obvious way). 3. Related to point 2, the way ASTRAL computes branch lengths is *very* dependent on ILS assumptions. So, you should probably ignore branch lengths. Ditto for localPP. You don't want to use them. 4. If you run ASTRAL with -t 1 it gives you quartet scores for each branch. The quartet score gives you a measure of the amount of discordance with each branch among your input trees. This is likely going to be useful for purposes. Thanks Siavash On Wed, Sep 26, 2018 at 12:18 PM Daniel Lang ***@***.***> wrote: > Hi Siavash, > > thank you for your response. I'm using astral to generate a consensus tree. > > Does this makes sense? As I wrote, topologies look great, although some > branch lengths appear longer than in other approaches I applied to the same > data... > > The data set liekly contains a lot of reticulation introgression events. > > Best, > Daniel > > Am 26. September 2018 20:34:07 MESZ schrieb Siavash Mirarab < > ***@***.***>: > >Hi Daniel, > > > >I hadn't seen the RRHS approach before and I am just starting to read > >it to > >understand how it works. So the short answer is I am not sure. > > > >Can you tell me how you are using RRHS with ASTRAL though? Are you > >using it > >to combine trees from multiple random selections (e.g., instead of the > >majority rule consensus in the original paper)? > >Or are you somehow making gene trees? > >In other words, is each input tree to ASTRAL estimated from data coming > >from the whole genome or is it from an individual locus? > > > >Thanks > >Siavash > > > >On Tue, Sep 25, 2018 at 12:17 PM Daniel Lang ***@***.***> > >wrote: > > > >> Ah, forgot to add that the resulting tree topology and branch lengths > >in > >> my data appear to make (biological) sense... > >> > >> — > >> You are receiving this because you are subscribed to this thread. > >> Reply to this email directly, view it on GitHub > >> > ><#38 (comment)>, > >or mute > >> the thread > >> > >< > https://github.com/notifications/unsubscribe-auth/AAybuGlKoGMPRQ34QvyAWMYTVw027xgDks5ueoFjgaJpZM4W5ToZ > > > >> . > >> > > > > > >-- > >Siavash Mirarab > > > > > >-- > >You are receiving this because you authored the thread. > >Reply to this email directly or view it on GitHub: > >#38 (comment) > > -- > Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail > gesendet. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#38 (comment)>, or mute > the thread > <https://github.com/notifications/unsubscribe-auth/AAybuItGBRlXK0DEAWXX1EHK2K7gYGy4ks5ue9LwgaJpZM4W5ToZ> > . > -- Siavash Mirarab -- You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: #38 (comment)

-- Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.