I was trying to make a tree/reconstruction with ML and Bayesian inference (BI) trees. I can make them individually but still don't know how to make/reconstruct trees. Could you please explain? Thanks!
I dont understand whats your question, sorry...
But to make a phylogenetic tree you need to understand few things...
The ML tree is a natural logarithm of the verisimilitude and it not represent the probability that the tree is correct, BUT is defined like the conditional probability to observe the data made by the hypothesis.
So, what Im saying? you need to considere few things to get your best hypothetic tree, the topology of the tree, the length of the branches and the parameters of substitution model, so I suggest first a NJ tree, so there you will estimate the topology and the length of the trees and then a ML (based on the NJ tree you got) to get the substitution parameters, with this, you will have the best tree.
How this work? well, the NJ tree you add is a "seed tree" this is a "base tree" so the program will start with this one to find the way to get the best option for you, more seeds you have, easier and better trees you will find, because with 10 sequences you will have millions of millions options, so the program can't get that quantity of options (unless you have a supercomputer).
I suggest you to learn how to use the software MrBayes (so you can get BI trees too), or get some help from someone who already know about this program, is very good and easy to use (if you know some about programming). In the other hand, you can use another kind of programs without the programming thing.
Hope this helps :)
thanks for your answer.
"so I suggest first a NJ tree, so there you will estimate the topology and the length of the trees and then a ML (based on the NJ tree you got) to get the substitution parameters, with this, you will have the best tree"
this i didn't get! i can make NJ tree and then can make ML also but it will be different tree. but how i merge 2 trees together? i use MEGA5 for making tree for now. thanks
Make a NJ tree first and then, make a ML using the tree you get from NJ, so the options that the software use to make the ML tree, will be better than just making the ML tree directly... I think there is an option in MEGA but sincerely I don't use MEGA because is not a valid software in a lot of journals, so I prefer to use other softwares. Im sorry :(
By the way, there is a better version of MEGA (MEGA 6.0) and there, the option (to use a tree to make another trees) is easier to find (really... really... I don't know where is that option!! :S )
Good luck!
It is very time consuming, but I think you should try Mr.Bayes (opensource software).
The first time I had to generate a tree I used MEGA 6.0 too, but the best method there is the ML, and nowadays most of the published trees follows a Bayesian approach.
I tried Mr.Bayes, following the manual as I am a self learner, and got a wonderful result. In my case the outcome was not so different topologically from the one generated in MEGA, but Mr.Bayes provides a nexus and newick formal results that is essential for other phylogenetic programs (as HyPhy).
I think you should try with Mr.Bayes, definetively. Just follow the tutorial/manual, making the necessary correction according to your own gene!
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts