The negative branch lengths may appear when there are distances in the distance matrix that are incompatible with the tree, i.e. there is triangle inequality. One simple algorithmic fix is to display one (or more) branches with a negative branch length. This happens under different scenarios, as exemplified by others here. It is not always an indication of a problem with the data. For instance, if your sequences have multistate characters this happens frequently. Again, it is simply a mathematical trick to make the tree from incompatible distances. Use a ML tree with a proper model if you don't like it, and that tree will likely be better anyway.

Did you use NJ method? the link might be useful

http://research.amnh.org/~siddall/methods/day6.html

Dear Dr. Patel,

Would you mind giving some more details on what have you done?

Cheers

Negative branch lengths do not arise frequently and only happen with data which is difficult to align correctly. But, I never had such experience.

You can read this for help. http://mikrobiologie.uni-graz.at/public/GCG/seqweb_help/growtree.html

Dear Priyanka,

You are saying that you use different length of sequences. If you want to have correct result in phylogeny the proper alignment is the most important prerequisite. Try to align the sequences by cutting protruding edges to have equal length of all sequences at both ends and remove all gaps in the alignment program and then try again with MEGA.

Good luck. Ivo.

Hi, I had such experience, you should check your alignment , if possible run it again to align properly. Another important step is to use sequences that are biologically similar. In your BLAST search consider E-value score of less than 10 to the power of minus 3 and identity ≥70%.

Regards

They can occur, but usually are an indication of a data set that has some kind of a "problem". Its not necessarily your fault. Its is very very unlikely that this is a problem of the software!! What are the pairwise distances of the sequences? If they are too large you should do some investigations.

No

Negative branch lengths can result if the sequences are of different lengths as mentioned previously but they can also arise if the sequences aligned are of very distantly related proteins (>30% homology). Avoid using very distantly related sequences

I have never had that problem with MEGA or other phylogenetic software. Realign your data, make sure you are comparing the same proteins and choose and adecuate outgroup

The negative branch lengths may appear when there are distances in the distance matrix that are incompatible with the tree, i.e. there is triangle inequality. One simple algorithmic fix is to display one (or more) branches with a negative branch length. This happens under different scenarios, as exemplified by others here. It is not always an indication of a problem with the data. For instance, if your sequences have multistate characters this happens frequently. Again, it is simply a mathematical trick to make the tree from incompatible distances. Use a ML tree with a proper model if you don't like it, and that tree will likely be better anyway.

See this article for another explanation and fix

http://boopsboops.blogspot.com/2010/10/negative-branch-lengths-in-neighbour.html

The best fix is to use maximum likelihood (ML) or a Bayesian method to infer your tree. You can always "improve" on your NJ tree by using it as a starting tree in a ML search (just remove the "-" signs). For example, you can do that with PhyML:

http://www.atgc-montpellier.fr/phyml

By the way, PhyML will always start with an NJ tree (BioNJ) by default and will improve on it. BUt you could fix the tree topology and just reoptimize the branch lengths - you should never get negative values back (unless there is a bug in the program, which you should then report :-))

I do phylogenetic tree by MEGA. at Alignment step, it has Error: All sequences are of length zero (see attached photo). How can i solve it?