Verification of the partitioned MSA generated by using SeqGen

[lutteropp]

Because of problems that emerged with the "small_network" experiments (NetRAX inferring a 0-reticulation network that has better loglikelihood than the "true" simulated network on the simulated MSA data), I have decided that we need to double-check the simulated MSA.

Proposed verification experiment A:

1. Use raxml-ng on each partition of the MSA generated by seq-gen
2. For each raxml-ng-on-partition-belonging-to-a-displayed-tree run, check if raxml-ng indeed inferred that displayed tree.

Proposed verification experiment B:

1. Create MSAs for each displayed tree in separate SeqGen calls.
2. Merge these MSAs into one big partitioned MSA by hand.
3. Re-run NetRAX on such a dataset.
4. Verify that again the BIC sees no supporting signal for reticulations.

[stamatak]

very good verification setup

…

On 01.12.20 16:21, Sarah Lutteropp wrote: Because of problems that emerged with the "small_network" experiments (NetRAX inferring a 0-reticulation network that has better loglikelihood than the "true" simulated network on the simulated MSA data), I have decided that we need to double-check the simulated MSA. Proposed verification experiment A: 1. Use raxml-ng on each partition of the MSA generated by seq-gen 2. For each raxml-ng-on-partition-belonging-to-a-displayed-tree run, check if raxml-ng indeed inferred that displayed tree. Proposed verification experiment B: 1. Create MSAs for each displayed tree in separate SeqGen calls. 2. Merge these MSAs into one big partitioned MSA by hand. 3. Re-run NetRAX on such a dataset. 4. Verify that again the BIC sees no supporting signal for reticulations. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#21>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGXB6QF7N3EWJ6NMUWWOTLSST3YNANCNFSM4UJC77JA>.

-- Alexandros (Alexis) Stamatakis Research Group Leader, Heidelberg Institute for Theoretical Studies Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology www.exelixis-lab.org

[celinescornavacca]

other suggestion: take a network with one reticulation and inheritance probs of around 0.5 for each of the reticulated branches and generate a MSA of length 1000, 2000 etc and see how the BIC evolves

[stamatak]

That's a good additional approach, I would not disregard the other two though ...

…

On 02.12.20 10:42, celinescornavacca wrote: other suggestion: take a network with one reticulation and inheritance probs of around 0.5 for each of the reticulated branches and generate a MSA of length 1000, 2000 etc and see how the BIC evolves — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#21 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGXB6W5SRK7WG3DAEKBHLTSSX4WNANCNFSM4UJC77JA>.

-- Alexandros (Alexis) Stamatakis Research Group Leader, Heidelberg Institute for Theoretical Studies Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology www.exelixis-lab.org

[celinescornavacca]

Yep, it is an additional one, if it was not clear