Issue with Lambda Structure Specification for Species Tree (sp1:30.4,(sp2:20,(sp3:11.9,(sp4:5.6,sp5:67):8.0):31):23.4)

[Fantacooke2]

Dear CAFE5 Development Team,

I am writing to request assistance regarding an issue with specifying the lambda structure for my species tree in CAFE5.

My species tree is provided as follows (with branch lengths):
  (sp1:30.4,(sp2:20,(sp3:11.9,(sp4:5.6,sp5:67):8.0):31):23.4);

According to the manual, the lambda structure should use the same NEWICK tree structure as the tree command but exclude branch lengths and substitute integer values (ranging from 1 up to n, where n is the total number of branches) in place of branch lengths. Matching integers indicate that those branches share the same λ value.

Based on this guideline, I removed the branch lengths from my species tree to obtain the topology:
  (sp1,(sp2,(sp3,(sp4,sp5))));
Then, assuming a fully bifurcating tree with five taxa (which should have 2×5–2 = 8 branches), I attempted to assign the same lambda class (λ = 1) to all branches by constructing the following lambda structure:
  ((1:1,1:1):1,(1:1,(1:1,1:1):1):1);

However, when running CAFE5 with this lambda structure, I encounter the following error:
  "ERROR(lambda -t): Branch lambda classes not totally specified.
  (1
  You have to specify lambda classes for all branches including the internal branches of the tree.
  There are total 8 branches in the tree.
  std::exception"

I have double-checked that my lambda structure should indeed reflect 8 branches, and yet the error persists. Could you please advise if I have misinterpreted the guidelines for constructing the lambda structure? Is there any nuance regarding the use of the branch lengths or the precise syntax (for example, retaining any aspect of the original NEWICK structure) that I might be overlooking?

I appreciate your time and assistance with this matter and look forward to your guidance.

Best regards
Fantacooke

[benfulton]

Your lambda tree does not match the structure of your species tree. Where you have sp1 in your tree you have a pair of nodes in your lambda tree. But also, you've renamed all of the species in your tree to 1. The structure is generally <species_name>:<lambda>, so a tree that will match your species tree would be

(sp1:1,(sp2:1,(sp3:1,(sp4:1,sp5:1):1):1):1);

Note also that you've only specified a single lambda here. Using this tree won't work any differently than not specifying a lambda tree at all.

[Fantacooke2]

Thank you very much for your feedback, @benfulton. I tried using the lambda
(sp1:1,(sp2:1,(sp3:1,(sp4:1,sp5:1):1):1):1);
but I received the same error. I ran CAFE with the lambda -s option (which was successful) instead of using the lambda -s -t (sp1:1,(sp2:1,(sp3:1,(sp4:1,sp5:1):1):1):1); command that I intended to use. In fact, if the tree string had worked successfully, I planned to assign a lambda value of 2 to all species except sp1—since sp1 belongs to a different family—and thus estimate multiple lambda values rather than a single lambda.

[benfulton]

That should work if you have everything set up correctly. I ran it like this:

$ cat tree.nwk 
(sp1:30.4,(sp2:20,(sp3:11.9,(sp4:5.6,sp5:67):8.0):31):23.4);
$ cat lambda_tree.nwk 
(sp1:1,(sp2:1,(sp3:1,(sp4:1,sp5:1):1):1):1);
$ cafe5 -i results/simulation.txt -t tree.nwk -y lambda_tree.nwk 

Command line: ../build/cafe5 -i results/simulation.txt -t tree.nwk -y lambda_tree.nwk 
WARNING: Tree is not ultrametric

WARNING: No root family size distribution specified, using uniform distribution
Searching for 1 lambdas


Optimizer strategy: Standard Nelder-Mead

[Fantacooke2]

Thank @benfulton . It is now running, but I guess it will run through because I have over 20 k families despite that I filter out family with gene copy less than 100 (as suggested in previous CAFE version tutorials). I read some of the discussion/issues about CAFE5 and it seems it does not support such a large dataset. I will maybe remove more of genes or some of the species and see how it goes. On a related note, if I manage to bring the number of gene family down to let's say 500, and want to use multiple lambda, the tree can modified to (sp1:1,(sp2:2,(sp3:2,(sp4:2,sp5:2):2):2):2); assuming that sp1 has a lower evolution rate (and belong to another family actually)?

[benfulton]

Yes, that would give you a separate lambda calculation for sp1.