The calculation of a phylogenetic tree does not depend from bootstrapping. These are two separated calculations, even though some software programs do link both calculations in one run.
There are many ways to calculate a phylogeny. You can use the "genetic distance" (if your data are sequences) in a neighbor-joining analysis. This is sparsely memory consuming method, but not very accurate (e.g. SplitsTree: http://ab.inf.uni-tuebingen.de/software/splitstree4/). Another possibility is to calculate a "parsimonic" phylogeny. This method is based on the assumption that evolution "went" the shortest way. Parsimony has been heavily criticized and becoming less and less popular. Finally, you can apply "maximum likelihood" that is a statistical calculation of the most likely tree based on different evolutionary models. That's what I would recommend you to do. You can run, for example, RAxML, PhyML, Garlic, MEGA etc. Or, alternatively, run a Bayesian analysis. MrBayes: http://mrbayes.sourceforge.net/
But you need to find out the substitution model that fits best to your data. You can find it out using jModelTest (former name just ModelTest) from David Posada's bioinformatic lab (http://darwin.uvigo.es/our-software/).
Finally, bootstrapping or not? Yes, do bootstrapping. The quality of a phylogenetic tree is usually measured by statistical bootstrapping. The bootstrap is a computer-based technique for assessing the accuracy of almost any statistical estimate. Felsenstein (1985; Evolution 39:783–791) introduced the use of the bootstrap in the estimation of phylogenetic trees. His technique provides assessments of “confidence” for each clade of an observed tree, based on the proportion of bootstrap trees showing that same clade.
In practice this means:
- look for the option "bootstrapping" in your software
- apply at least 100 replicates; much better would be more than 100, something between 500-1.000 replicates, but this depends on your computer working capacity
- check the resulting tree from bootstrapping with the phylogenetic tree you have produced: do you obtain the same clades? If yes, check the bootstrap % that you obtained for the different clades.
- if you have applied a maximum likelihood method to calculate the phylogenetic tree: bootstrap values 70% are statistical significant; the higher the bootstrap value >70% the better the confidence in that clade.
Thank You Carolina ..for the detailed explanation ..i have further question like if the bootstrap value is below 50% for some branches in a clade not all ..can we change the values ? or in that case we have to again run in some other software ? What will be the plausible solution ? As i was constructing a tree in Raxml by ML method and i saw some values after bootstraping 1000 replicates was less than 50 ..thn what can be done ?
You are not allowed to change results in any way. But the analysis may not be optimal.
1) Data: Are you sure your sequence data contain enough phylogenetic information? How long is your alignment? One or more loci? How many informative sites does the alignment contain?
2) Biology: How closely/distantly related do you expect are your taxa? If they are very closely related, may be that only 1 locus alone does not contain enough phylogenetic information to separate them in different clades. In this case, you will need several loci to unravel the relationship.
3) Analysis: RAxML is very fast and a good tool. But it runs data always with the GTR substitution model. Maybe that GTR is not the appropiate evolutionary model for your data. You can check the model using jModelTest. If you are not familiar with these computer programs, you can use PhyML including model search on the South French online bioinformatic platform: http://www.atgc-montpellier.fr/phyml/. Use "Automatic model selection" and "AIC". Later, use "SPR" and "number of starting trees" "no". Later, use 100 bootstraps, for example, to see quickly the results. You will obtain an e-mail with results or you can download them from the atgc-platform. Check the log-file: - model suggested; - how many informative sites does the alignment contain?
Finally, a combination of the different problems from 1) - 3) may distort your results. Anyway, 50% is too less support to pretend that you have different species.