I have a small data set consisting of 16 sequences, 919bp each. I am trying to determine the phylogenetic relationship among the individuals. I am wondering if the neighbor-joining method, maximum likelihood or bayesian analysis works best for small data sets like mine. Thank you.
Do you have any sequence to root your analysis (or does any of those sequence is a root independent of your samples)?
In case the answer is yes, I may suggest you to make an exploratory analysis via PHYML: http://www.atgc-montpellier.fr/phyml/ using the presets of all the features. If you want something more used in papers, go for RAxML. The CIPRES web server has the newer version.
What is the question/purpose of your analysis? If Maximum Likelihood answers it, there is no need to look for something bayesian. And if it is necessary, you can give it a try (equally exploratory) via MrBayes in https://ngphylogeny.fr/tools/tool/281/form
In the case the answer is no, I completely suggest you the SPLITSTREE software. (http://www.splitstree.org/), its very helpful to resolve phylogenetic relations between individuals inside the same group and it has various algorithms to construct different trees. Is super user-friendly, and due the size of your data, it will work very fast.
I truly recommend this crash crash course (even though it is meant for virology, it results very helpful to understand phylogenetic principles): https://viralzone.expasy.org/e_learning/phylogenetics/content.html
In addition to the above answers, you should download e.g. jModeltest to determine the best fit model. Load your data, compute likelihood scores and do AIC calculations to find the best model for Maximum Likelihood analyses and BIC calculations for Bayesian analyses. Since you want to do phylogenetic analyses, neighbor-joining is rarely a suitable method.
I would suggest using MEGA software for such small datasets.
https://www.megasoftware.net/
I think it is better to test at least three different building methods-NJ, ML, ME, if the analyses using three different building methods agrees then most probably the analysis is based on the sequence and not biased on the building method employed.
Hi all,
Thank you for your responses, I found them all to be helpful. I applied NJ, ML and Bayesian analyses using an out group to root the tree. The results showed the same tree configuration for all. I got strong posterior probability values for the Bayesian analysis (98-100 for most nodes), but I got lower bootstrap support values in the ML tree (61-72). Is this due to the different nature of the analyses?
Which software did you use for each analysis, Emma Larkin-Gero?
Definitively, rooting helps a lot. In the case of ML and bayesian differences, they may be a result of the bootstrapping and MCMC parameters, respectively. My best suggestion is too look up for papers that present analysis and results the most similar (in methodology and concept) of what you are trying to accomplish, and make an idea of the best workflow to give robustness to your results.
Jose Alberto Lopez-Aleman Thank you for your advice. I used MrBayes, Garli for ML and MEGA for NJ.
- Badges
- Science topic
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics