I have a sequence of 16srRNA, having analyzed it with BLAST I came up with a genus: Kocuria. I picked 5 sequence of highest identity and lowest e-value of my blast search plus 15 16s gene from another Kocuria species randomly. I picked Sinomonas albida strain LC13 as outgroup. (kocuria and sinomonas belong on Micrococcaceae). I constructed a phylogenetic NJ phylogenetic tree using mega 6 and set bootstrap replications value of 1000. Then tree I obtained have really low bootstrap value, here is the picture of my result. Where did I do wrong, how can i make a better tree?
Low bootstrap values are typically the result of insufficient data (i.e. short sequences) but there may be other more general problems such as an inappropriate substitution model or suboptimal sequence alignment. My suggestions are (i) run the alignment again and check the quality of the GenBank sequences, (ii) check if you used the most appropriate substitution model, (iii) use a different tree-building algorithm (ML or Bayesian analysis, or both). If the support levels remain low, you really need more data, i.e. longer sequences, and preferably a few additional makers.
Low bootstrap values are typically the result of insufficient data (i.e. short sequences) but there may be other more general problems such as an inappropriate substitution model or suboptimal sequence alignment. My suggestions are (i) run the alignment again and check the quality of the GenBank sequences, (ii) check if you used the most appropriate substitution model, (iii) use a different tree-building algorithm (ML or Bayesian analysis, or both). If the support levels remain low, you really need more data, i.e. longer sequences, and preferably a few additional makers.
With your scale bar of 0.005 indicating 5 changes per 1,000 bases and the typical Kocuria 16S rRNA sequence in GenBank not much more the 1,000 bases in length, I would expect bootstrapping with that many species/isolates to give this result. I doubt there is anything wrong. If you now choose just two isolates from each of the clades of interest to you, and built a new tree , I suspect you could show high values. But I am not sure if a small tree with high bootstrap values is more informative than a large tree with many isolates and low values.
Agreed with Brian, the tree looks very typical for trees built based on such datasets. One of the tricks, which can help to 'boost' your bootstraps a little bit, is the good/robust outgroup for the tree, but it won't bring values much higher.
I agree with the others, It seems that you need more appropriate markers beside the current data, and its better to construct phylogenetic tree using different sort of parameters such as the most probabilistic methods e.g ML/BI approaches.
I agree with others - to get more resolution you need more variable positions. That can be achieved by incorporating more sequence data, for example add 23S rRNA too. make concatenated alignments from 16S and 23S. Assuming the same variation as in 16S it should add approximately 15-20 variable positions (to those what you already have in 16S).
In addition, when you are trying get resolution among closely related bacteria (strains) it might be good idea to pick sequences from the same rRNA operon (Kocuria marina have 3 rRNA operons). For some bacterial strains distance between 16S rRNA from different operons of same strain might be bigger bigger, than differences between the sequences from the same operon of different strains. - i.e. variation inside genome exceeds variation between genomes and you results will depend from 16S rRNA sequences you have happened to pick up. But is it you case, you need to test first.
- Badges
- Science method
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics
- Phylogenetic Analysis