Suggested commands for huge 244 taxa/126 characters matrix

[Geraldo Alonso Góes Rocha]

Hello!

I'm having a challenging time with my matrix. I'm a master's student in Brazil with very limited time to achieve some results. I'm working with a modified version of Davis (2011) baridinae matrix to include my studied taxa. The problem is since the matrix is huge, it's taking days for the hits to occur with the current commands I'm setting for the program to run. I'm currently using these commands: 

hold 99999

collapse 3

sec : slack 3 xss 3+3-1 gocomb 10 combstart 5 fuse 3 drift 6 ; drift : rfit 0.10 num 150 nogiveup ; xmult = level 10 hits 200 rep 5 drift 20 fuse 6 gfuse 4

these commands were adapted from Ramírez (2014) adapted to suit my matrix requirements after errors started to show

my advisor says it's important for us to keep the 200 hits setting. The best score I've achieved so far is 3317, but after about 150 hours I'd had only 6 hits. My time is running out and I can't find a better-optimized way to search. I have barely any knowledge of the program commands, and I don't know where to begin, or which numbers are best for each algorithm, so I was hoping someone could help me solve this issue

[Ambrosio Torres Galvis]

Dear Geraldo, 
I forgot to tell you that if you have more than 100 taxa, it is better to also include sectorial searches, so you don't need these arguments: "nocss noxss norss".
Instead, you can include all sectorial searches using:  "css xss rss"
Use "help xmult;" and "help sectsch;" to obtain more information. 
Best,
--Ambrosio T.

El jue, 2 ene 2025 a las 13:30, Ambrosio Torres Galvis (<atorre...@gmail.com>) escribió:

Dear Geraldo, 
the routine you are using seems unnecessarily heavy. One of the main reasons TNT is widely regarded as the most efficient parsimony program (e.g., Goloboff et al., 2022) is due to its highly optimized algorithms. This efficiency means you don’t need overly complex routines to find the Most Parsimonious Trees (MPTs).

I respectfully recommend using a more efficient approach. For instance, here’s the routine we used to analyze 157 phylogenomic datasets (Torres et al., 2021):

xmult= rep 3 hits 3 fuse 10 ratchet 10 drift 10 nocss noxss norss; bbreak;

 Alternatively, you can use the scripts provided in Torres et al., (2022). While primarily intended for phylogenomic datasets, they are also suitable for morphological matrices.
Of course, your matrix could have special subtleties that make you feel you need deeper searches; however, I think you will be fine using much lighter searches.
If you’re still uncertain about finding all the MPTs, you can gradually increase the parameter values. You’ll likely find that the results converge to the same trees quite quickly.

All the best,
-- Ambrosio T.

--
Has recibido este mensaje porque estás suscrito al grupo "TNT-Tree Analysis using New Technology" de Grupos de Google.
Para cancelar la suscripción a este grupo y dejar de recibir sus mensajes, envía un correo electrónico a tnt-tree-analysis-using-n...@googlegroups.com.
Para ver este debate, visita https://groups.google.com/d/msgid/tnt-tree-analysis-using-new-technology/cfb8dc70-78e5-4cd0-8318-52668d26af49n%40googlegroups.com.

--

Ambrosio Torres G.

- Researcher / Investigador [Ctr. for Integr. Biodivers. Discov. - Museum für Naturkunde, Invalidenstraße 43, 10115 Berlin, Germany]

- Ph.D. Biol. Sci. / Dr. en Cs. Biol.  [UNT, S. M. de Tucumán, Argentina]

- B. Sc. / Biólogo [UIS, Bucaramanga, Colombia]

• Institutional email: Ambrosio.T...@mfn.berlin
• Google Scholar: https://scholar.google.com/citations?user=UshnduEAAAAJ&hl=es
• ResearchGate: https://www.researchgate.net/profile/Ambrosio_Torres_G
• GitHub: https://github.com/atorresgalvis

--

Ambrosio Torres G.

- Researcher / Investigador [Ctr. for Integr. Biodivers. Discov. - Museum für Naturkunde, Invalidenstraße 43, 10115 Berlin, Germany]

- Ph.D. Biol. Sci. / Dr. en Cs. Biol.  [UNT, S. M. de Tucumán, Argentina]

- B. Sc. / Biólogo [UIS, Bucaramanga, Colombia]

• Institutional email: Ambrosio.T...@mfn.berlin
• Google Scholar: https://scholar.google.com/citations?user=UshnduEAAAAJ&hl=es
• ResearchGate: https://www.researchgate.net/profile/Ambrosio_Torres_G
• GitHub: https://github.com/atorresgalvis

[Ambrosio Torres Galvis]

Dear Geraldo, 
the routine you are using seems unnecessarily heavy. One of the main reasons TNT is widely regarded as the most efficient parsimony program (e.g., Goloboff et al., 2022) is due to its highly optimized algorithms. This efficiency means you don’t need overly complex routines to find the Most Parsimonious Trees (MPTs).

I respectfully recommend using a more efficient approach. For instance, here’s the routine we used to analyze 157 phylogenomic datasets (Torres et al., 2021):

xmult= rep 3 hits 3 fuse 10 ratchet 10 drift 10 nocss noxss norss; bbreak;

 Alternatively, you can use the scripts provided in Torres et al., (2022). While primarily intended for phylogenomic datasets, they are also suitable for morphological matrices.
Of course, your matrix could have special subtleties that make you feel you need deeper searches; however, I think you will be fine using much lighter searches.
If you’re still uncertain about finding all the MPTs, you can gradually increase the parameter values. You’ll likely find that the results converge to the same trees quite quickly.

All the best,
-- Ambrosio T.

El mié, 1 ene 2025 a las 22:07, Geraldo Alonso Góes Rocha (<geraldo.g...@hotmail.com>) escribió:

--
Has recibido este mensaje porque estás suscrito al grupo "TNT-Tree Analysis using New Technology" de Grupos de Google.
Para cancelar la suscripción a este grupo y dejar de recibir sus mensajes, envía un correo electrónico a tnt-tree-analysis-using-n...@googlegroups.com.
Para ver este debate, visita https://groups.google.com/d/msgid/tnt-tree-analysis-using-new-technology/cfb8dc70-78e5-4cd0-8318-52668d26af49n%40googlegroups.com.