How to make or reconstruct a phylogeny tree like above? Is it possible to make ML/NJ tree by MEGA or which software can make one?
There are four things you should bear in mind: 1) what you are trying to do is essentially a statistical inference: we just don't know the history of life, so we try to make the best approximation from the data we have; 2) what you expect is a phylogenetic tree, which is a geometrical depiction of such history (a caricature if you wish) and has geometric properties that are relevant to compare it with similar trees; 3) you need to sort out what information is useful (i.e. phylogenetic signal) and what is basically noise (distorting that signal); and 4) beware of artifacts that are unavoidable and can be quite misleading. Of course you are not expected to do this at once, but these points are most important when you are to decide what algortithm yuou want to use. So, distance-based methods such as NJ may be popular among microbiologists for example, but hte are not phylogenetic inferences --just similarity clustering. When using DNA data, a further issue is most relevant: given that you have only four possible character states (four possible nucleotides in each position), reversals are expected over time, thus confusing or even erasing the phylogenetic signal. That is why you need to have a model of evolutionary change. So, here is what I woud recommend: use MODELTEST or a similar program to find out what mdel best fits your data. Then go to www.phylogeny.fr and paste your sequences there. It' very user-friendly, reliable, fast and free. Make sure you select the adequate evolutionary model. The default method is a powerful maximum-likelihood method --trust it. The tree you get can be shown in different ways. Then read about what the different steps of the process are, and get familiar with what the machine did. You will have a reasonable understanding and a robust tree.
There are four things you should bear in mind: 1) what you are trying to do is essentially a statistical inference: we just don't know the history of life, so we try to make the best approximation from the data we have; 2) what you expect is a phylogenetic tree, which is a geometrical depiction of such history (a caricature if you wish) and has geometric properties that are relevant to compare it with similar trees; 3) you need to sort out what information is useful (i.e. phylogenetic signal) and what is basically noise (distorting that signal); and 4) beware of artifacts that are unavoidable and can be quite misleading. Of course you are not expected to do this at once, but these points are most important when you are to decide what algortithm yuou want to use. So, distance-based methods such as NJ may be popular among microbiologists for example, but hte are not phylogenetic inferences --just similarity clustering. When using DNA data, a further issue is most relevant: given that you have only four possible character states (four possible nucleotides in each position), reversals are expected over time, thus confusing or even erasing the phylogenetic signal. That is why you need to have a model of evolutionary change. So, here is what I woud recommend: use MODELTEST or a similar program to find out what mdel best fits your data. Then go to www.phylogeny.fr and paste your sequences there. It' very user-friendly, reliable, fast and free. Make sure you select the adequate evolutionary model. The default method is a powerful maximum-likelihood method --trust it. The tree you get can be shown in different ways. Then read about what the different steps of the process are, and get familiar with what the machine did. You will have a reasonable understanding and a robust tree.
Thanks Cristian. I have tried to do it in www.phylogeny.fr but program showing that data/sequence is more so it can't process. but my sequence is only 62!
I also tried it with 120 sequences and it is working perfectly. I used NEXUS format file.
Hi Ahasan,
The problem must be in the format. There are some sequences appearing as example, in the box where you have to put your data. Make sure your sequences look just like the example. And also, make sure there are no line returns within any sequence.
In case you want to use a Bayesian method, the limit in this platform is 30 sequences (I think). However, I don't recommend MrBayes, because it easily yields artifacts such as long-branch attraction.
Hope this helps! If you have any trouble, let me know.
thanks. please check in your inbox @Cristian. still can't construct a tree inference with 2 methods (i.e NJ/ML or ML/BI). thanks
- Badges
- Science topic
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics