Allow to create multifurcated trees

[westram]

Shall allow to change branchlengths in a given according to existing bootstrap values.

Input

• (current)
• max. bootstrap value (MBS)
• max. branchlength (MBL)

Modifies all branches with bootstrap⇐MBS and length⇐MBL, such that their length and bootstrap will be set to zero.

The removed branchlength shall be distributed to the child-branches of the removed branches. This is not correct, but practible.

A correct branchlength distribution for the removal of branch X in

A       C
 \_____/
 /  X  \
B       D

would need to keep all 6 distances between (A,B,C,D) somehow constant (at least relatively). This is quite difficult to obtain.

Discussion with WL revealed that such a distribution already was in ARB @ its beginnings (applied to manual branchlength modification via display) and has been removed, since it proved impractical, at least for manual branchlength modification. Looking at the above as dendrogram may e.g. look like:

+------A
|
|   +----B
+---+
    |   +---C
    +---+
      X +---D

Removing branch X with "correct distribution" will cause "parent" branches to change their length.

If @ all, implement "correct distribution" as option.

[westram]

preparative work: log:branches/tree@11425:11449

[westram]

implemented basic functionality with [11466]

[westram]

Replying to westram:

The removed branchlength shall be distributed to the child-branches of the removed branches. […] Removing branch X with "correct distribution" will cause "parent" branches to change their length.

Decided to implement a proportional length distribution. The length of each eliminated branch gets added to it's "adjacent branches", where longer branches receive more length than shorter branches and branches with length zero never receive additional length. They build a multifurcation and the branches beyond the zero-length-branches will receive part of the distributed length.

TODO: Length distribution has to take place after deciding how much length will be distributed to each branch.

[epruesse]

We also need to have a proper documentation of the algorithm, both for the paper and in the help file.

[westram]

Replying to westram:

TODO: Length distribution has to take place after deciding how much length will be distributed to each branch.

implemented by [11478]

[westram]

[11483] adds a MULTIFURC mode to display.

new TODO

• avoid that negative branchlengths are generated by 'multifurcate()'
• check strange problems @ root node

[westram]

Replying to westram:

• check strange problems @ root node

fixed by [11484]

[westram]

• avoid that negative branchlengths are generated by 'multifurcate()'

fixed by [11485]

[westram]

by [11488]