I am trying to construct a phylogenetic tree between some protein sequence alignments. I am struggling however with the interpretation of which tree I should use as the program seems to make many iterations. I need to know what statistics are valuable in helping me decide which tree best represents the data. I have done both a Maximum-likelihood phylogenetic tree and a Bayesian tree as well. I am currently using the geneious prime software to construct my trees so anyone with experience in this database your knowledge would be super useful.
Phylogenetic tree construction can be a complex process, and there are several factors to consider when interpreting the results. The choice of statistical method (such as maximum likelihood or Bayesian) can influence the topology of the resulting tree, as can the choice of evolutionary model and the parameters used in the analysis.
One important statistic to consider when evaluating the quality of a phylogenetic tree is the bootstrap support value. This value represents the percentage of times that a particular branch is supported by resampling the data. Generally, a bootstrap support value of 70% or higher is considered to be significant support for a particular branch. You can also evaluate the posterior probabilities in a Bayesian analysis, which provide a measure of support for each branch.
The workflow I usually take is to build several trees with a low number of iterations or bootstrapping (to save time) from the same data input using different models. By comparing their topology, I get a reasonable insight into how robust the tree is. When almost all models result in a tree with similar topology, you can choose one of these methods that seems appropriate and rerun that method with more bootstraps or iterations for statistical support. One important factor to consider is the overall topology of the tree. If the tree has good support values for all major branches and is consistent with previous research on the organism(s) in question, it is likely a good representation of the data.
Another thing you could do if you have constructed multiple trees using different methods or parameters is to compare the trees using statistical tests such as the Shimodaira-Hasegawa test or the Kishino-Hasegawa test. These tests can help you determine whether one tree is significantly better than another based on the likelihood scores and the number of parameters used.
Finally, it can be useful to visually inspect the trees to look for patterns and outliers and compare this with the alignment. If there are a few sequences that seem to be clustering in unexpected ways or are not well-supported by the data, it may be worth investigating further to see if there are any errors or issues with the data. You can choose to remove an outlier to reduce its interference with correct tree building, and/or you can add other sequences which allows for clustering when a branch in the tree does not have many closely related species. A lack of sister species can result in a branch that starts apparing in unexpected places as it does not really have a clustered place in your tree.
I hope this helps! It usally takes quite some work to construct a good tree but eventually you will get there.
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Biomolecules
- Peptides