Hello Bioinformaticians
I'm using IQ-tree (Maximum-likelihood method) to construct phylogenetic trees for a set of protein sequences. The number of replicates (1000) for ultrafast bootstrap. So here I read a paper and It mentions that for ultrafast bootstrap, you should only start to believe in a clade if its support is > 95%. I have support like 65, 70 to 100. So my question : what is the cut-off for ultrafast bootstrap values?
There is no formal cutoff for what to believe with ultrafast bootstrap (UB). Some people will argue that 70% UB values are reliable, and some people will actually buy that argument. That said, the IQ-TREE authors have used and tested the program more than just about anyone else, so their recommendations matter. I believe they also recommend at least 1000 replicates for UB, while 100 may be enough for nonparametric bootstrap (NB). UB is not as rigorous as NB IQ-TREE reconstruction, and that's the price of speed.
If the branches with lower support are far away from the root and in parts of the tree with only a couple of leaves, there is no shame in leaving the tree as such and accepting that parts of it are less reliable. If the poor support branches are near the root (say, if they include half the leaves in the tree), then most people will have an issue with it. You can always try NB and see if that gives better support, though in my experience UB with 1000 replicates typically gives better branch support than NB with 100 replicates.
