Hi, I was using rapidNJ to perform neighbour-joining but it gave me a couple of nodes with negative branch length. Is there any way I can fix it? If rapidNJ won't do it, do you have any recommended program to do a better job? Thanks.
Negative branch lengths are possible in theory and might occur in the output of other programs (e.g. FastTree) as well. See the short section on negative branch lengths here for example. If your downstream analysis requires non-negative branch-lengths, a quick hack is to replace negative entries with zero (even though in theory you should add the difference to the adjacent branch length). The following should do (untested!):