Difficulty altering tree in HBL

[halabikeren]

Hey dear team,

I am trying to alter a tree in HBL (specifically, adding an internal node with a single child and assigning length to the branch whose upper edge is the new internal node).

Here is an example of how I'm trying to do so:

/* attempt to edit Tree instances */

/* toy example for usage: with branch lengths */
fprintf(stdout, "\n**** test with original branch lengths in tree ****\n");
Tree T1 = "((S1:0.1,S2:0.1)Node1:0.1,S3:0.1,S4:0.2);"; // maybe needs to be a topology instance - try Tree first
fprintf (stdout, "tree before edit: ", Format (T1,1,1), "\n");
T1 + {"PARENT" : "internal1" , "WHERE" : "S1", "LENGTH" : 0.42}; // branch lengths are not conserved
fprintf (stdout, "tree after edit: ", Format (T1,11,1), "\n");

/* toy example for usage: without branch lengths */
fprintf(stdout, "\n**** test with no original branch lengths in tree ****\n");
Tree T2 = "((S1,S2)Node1,S3,S4);"; // maybe needs to be a topology instance - try Tree first
fprintf (stdout, "tree before edit: ", Format (T2,1,1), "\n");
T2 + {"PARENT" : "internal1" , "WHERE" : "S1", "LENGTH" : 0.42}; // branch lengths are not conserved
fprintf (stdout, "tree after edit: ", Format (T2,1,1), "\n");

This yields:

**** test with original branch lengths in tree ****
tree before edit: ((S1:0.1,S2:0.1)Node1:0.1,S3:0.1,S4:0.2)
tree after edit: (((S1:0.1)internal1:-1,S2:0.1)Node1:0.1,S3:0.1,S4:0.2)

**** test with no original branch lengths in tree ****
tree before edit: ((S1:-1,S2:-1)Node1:-1,S3:-1,S4:-1)
tree after edit: (((S1:-1)internal1:-1,S2:-1)Node1:-1,S3:-1,S4:-1)

This is what I'm expecting to receive:

**** test with original branch lengths in tree ****
tree before edit: ((S1:0.1,S2:0.1)Node1:0.1,S3:0.1,S4:0.2)
tree after edit: (((S1:0.42)internal1:-1,S2:0.1)Node1:0.1,S3:0.1,S4:0.2)

**** test with no original branch lengths in tree ****
tree before edit: ((S1:-1,S2:-1)Node1:-1,S3:-1,S4:-1)
tree after edit: (((S1:0.42)internal1:-1,S2:-1)Node1:-1,S3:-1,S4:-1)

Am I not using this syntax correctly? Additionally, when adding a node with a name that has a label, the node name is reformatted. For example, node name I0{Test} is reformatted into I0_Test_ when added to the tree. Is there any way to avoid this issue, or alternatively, add the labels to the nodes after altering the tree?

Many thanks! Keren

[spond]

Dear @halabikeren,

The most general form of Tree + object will splice the target branch and like so:

Splice the target branch with existing length 

Before

       |-----L-----| [B] ...
 ...---|
       |--------- ...

T + {"WHERE": "B", "NAME" : "new_child", "PARENT_NAME" : "splice_parent", "LENGTH" : x, "PARENT_LENGTH" : y}

After:

       |-y-|-L-----| [B] ...
       |   |
       |   |--x---- [new_child]
 ...---|
       |--------- ...

Omitting NAME or PARENT_NAME (but not both!) will split the existing branch into two components (with a single node in between).

None of these operations will ever change the length of the exiting branch (L).

Best, Sergei

[halabikeren]

Thank you for the clarification, Sergei!

Additionaly, when I try to add a node with a name tthat includes a label, the node name is reformatted.

This example:

fprintf(stdout, "**** add a new parent of B which will be a child of A ****\n"); 
Tree T1 = "(((B:1,C:1)A:1),D:1);";
fprintf(stdout, "tree before edit: ", Format(T1,1,1), "\n");
T1 + {"WHERE": "B", "PARENT" : "parent{FG}", "LENGTH" : 0.42, "PARENT_LENGTH" : 0.042};
fprintf(stdout, "tree after edit: ", Format(T1,1,1), "\n\n");

Yiels:

**** add a new parent of B which will be a child of A ****
tree before edit: ((B:1,C:1)A:1,D:1)
tree after edit: (((B:1)parent_FG_:0.042,C:1)A:1,D:1)

Is there any way to avoid this re-formatting, or alternatively, add the labels to the nodes after altering the tree?

Thanks! Keren

[spond]

Dear @halabikeren,

HyPhy will not parse meta information when you call Tree + dict. In order to assign a model to a specific tree branch, please use something like the following code.

Model BSREL = (...);

...

SetParameter (T1.parent, MODEL, BSREL)

Also note that you can directly manipulate branch -> class (e.g. FG/BG) assignment as it is simply a dictionary.

Best, Sergei

[halabikeren]

Dear @spond,

Thank you for the great suggestion!

Since I need to convert the tree instance to a dictionary anyway, this could be a good solution for my case.

I found a way to do that using the function trees.ExtractTreeInfo():

LoadFunctionLibrary("libv3/tasks/trees.bf");
Tree T = (((A:1,B:1)C:2),D:3);
TStr = Format(T, 1, 1);
TDict = trees.ExtractTreeInfo(TStr);
fprintf(stdout, "TStr: ", TStr, "\n"); 
fprintf(stdout, "TDict: ", TDict, "\n");

This yields:

TStr: ((A:1,B:1)C:2,D:3)
TDict: {
 "string":"(A,B,D)",
 "string_with_lengths":"(A:1,B:1,D:3)",
 "branch length":{
   "A":1,
   "B":1,
   "D":3
  },
 "annotated_string":"(A,B,D);",
 "model_map":{
   "A":"",
   "B":"",
   "D":""
  },
 "partitioned":{
   "A":"leaf",
   "B":"leaf",
   "D":"leaf"
  },
 "model_list":{
  {""}
  }
}

Which is exactly what I need.

I have three questions regarding this procedure: 1) In the process of converting the string to a dictionary, the internal nodes appear to be lost. Is there any way to preserve them? 2) Allegedly, now I could alter the model_map and model_list entries in the dictionary, in order to add the labels. However, altering the "annotated_string" entry is a bit more challenging. Is it OK to ignore this entry, and rely solely on the model_map for the labeling (even though it would create a contradiction in the information of the annotated string and the model map)? 3) Since I'm already using the dictionary as the data structure representing the tree, can I "edit" branch lengths using the "branch length" entry in it, rather than editing them using string regex and sub-string replacement?

Thank you! Keren