low bootstrap value?

Question

[Kumar]

Hi all,

I have created a phylogenetic tree (with 500 bootstrap iteration) based on AMPHORA tool specified marker genes (28 marker protein sequences were extracted from 30 bacterial genome and concatenated, aligned by mafft and trimmed by gblock server). Even though the alignment is so good, but half of the clades bootstrap value is 10 to 20. I tried with various models, despite I could not get good bootstrap values (70 above). Is it possible to improve the bootstrap value in any way?

Thank you in advance.

[Joe]

Not really, short of adding more sequences AFAIK.

You shouldn't fixate on the bootstraps though. In most trees there will be some nodes with lower node supports, its fairly normal, particularly if your sequences are very closely related or very distant.

[Kumar]

@jrj.healey, As you mentioned, all my genomes belongs to the same species, varied by sub species level classification (pathovar). But, one thing I would like to ask you, the tree topology is slightly different from gyrb and 16s rDNA based tree of the same. Is it normal?

[Joe]

I would imagine thats somewhat to be expected, if you're typing with ~30 genes, thats a lot more information than just gyrB and 16S, so I wouldn't expect them to necessarily be identical, unless the 16S and gyrB phylogeny was perfectly correct from the beginning.

This assumes all else is equal though, and that there isn't something wrong with what you (or whoever did the 16S study) did.

[lieven.sterck]

In most cases you will (as jrj.healey points out as well) have to live with it.

Bootstraps can be "changed" by editing the input MSA for the phylogenetic analysis (more rigorous or more lenient editing).

You could also consider to switch to building the tree using the nucleotide alignment in stead of protein, in cases of high protein similarity/conservation it might help to give more resolution.

[Kumar]

@lieven.sterck, I have already done with nucleotide sequences, but end up with the same bootstrap values.

[Michael]

You could also try to create alignments for each protein first, and then concatenate the alignments after that. That might avoid artificial alignments across junctions. You could also try skip the gblocks trimming step, as it also might remove some phylogenetically important regions.

[Kumar]

@Michael Dondrup, As I already excluded gblocks trimming step, despite end up with the same bootstrap values. As you mentioned, I would do the individual protein alignment. Anticipated to obtain better bootstrap value.