Yes, you will have to align your outgroup sequence to the rest of your sequences. Any sequence based phylogenetic algorithm needs to know the inferred positional homology of the sequences to determine relationships. It doesn't matter whether it's a distance based tree, parsimony, max. likelihood, bayesian or whatever. The very first thing you need is an alignment of all sequences before any phylogenetic inference algorithm can start making inferences about the placements of the sequences in a tree.

And when you assign a sequence as an outgroup, you also impact inferences based on the positional homology of that outgroup's alignment with all the other sequences.

thanks. @Michael Black, let say my samples sequence bp is about 850bp (with forward and reverse after add together) and my out group sequence is about 800bp taken from NCBI. so when i add out group with samples sequence and do alignment and elimate the poor aligned position..i mean when became homologous then size is 659bp.

so is that alright or there would be any impact on analysis/results?

note: for that out group sequence, only 800bp is deposited in NCBI as partial sequence.

thanks so much!

That is the way I would do it, yes. For phylogenetic inference, you would use only the portion for which you have alignment across all the sequences. In a phylogenetic analysis, most algorithms will simply ignore missing data, so it does not contribute to or help in the assembly of the inferred tree. But, some algorithms also can be artifactually mis-led if there is a log of sequence with missing data (so long ends of non-aligned regions in many sequences, on one or both ends of your outgroup sequence).

So I would perform the alignment and make sure you have what you feel is the best possible overall alignment, then trim the ends to eliminate any large stretches of missing sequence.

Obviously internal gaps in the alignment are fine as those are actual inferred deletions or insertion events, but you don't want a lot of sequences in the alignment that have no actual alignment at the ends. Space filling the ends to match your longest sequence usually does nothing to help the phylogenetic inference, and with some algorithms, it can cause artifacts giving misleading trees.

thanks@Michael, so i can proceed with 659bp sequence length?? for your note, i did alignment by MAFFT and then elimate poorly sequence data by Gblock and then sequences shows homologous and all are in 659bp including out group sequence.

thanks,

And which software you use for phylogenetic analysis/to make tree and for other analysis? Thanks

Sounds alright to me. Now you just need to select your method of phylogenetic inference and build your tree(s).

thanks Michael. which software you use for phylogenetic analysis/to make tree and for other analysis? Thanks

There are plenty freely available. For most of my own work, I've used programs in the PHYLIP package - http://evolution.genetics.washington.edu/phylip.html

But the PHYLIP web site has links to a number of other good programs as well - http://evolution.gs.washington.edu/phylip/software.html

You should really decide on a method (Parsimony, Maximum Likelihood, Bayesian inference,...), then find a program. My past work used primarily Parsimony and Max. Likelihood methods, and I must admit I do not generally like distance based methods as they often are far too simplistic for inferring evolutionary phylogenies (but they make good exploratory tools as they are typically the least demanding computationally).

Thank you.

Hi Ahasan

All your questions have been very perfectly answered by Michael. I don't think I have anything extra to add. But just for the last part I would like to say something. There are many soft wares for building phylogenies. Many of them are freely available, almost all of them, except for PAUP, thats the only one I know which is a paid one.

PHYLIP is a good option as mentioned above. Then there is PAUP and WINCLADA which are parsimony based, WINCLADA is faster than PAUP. Then there is RAxML and GARLI, both likelihood based and both of which give good results and are being used currently.

Mr. Bayes, BEAST, *BEAST they all run on Bayesian framework. BEAST and *BEAST are used for dating the bayesian tree using a suitable calibration point relevant to your study.

Regards

Thanks @Kunal. i will have a look on those software you mentioned.

@Michael, could you please tell me, how to make a phylogenetic tree based on 3/4 methods which combine in a one tree. for example, i saw in one article they have show a tree "Consensus tree based on distance method, maximum likelihood, maximum parsimony and Bayesian inference methods..." if i want to make a tree like that, how and by which software i should use? thanks so much!

Hi Ahasan,

Although your question has been responded probably so far, you may be interested in knowing that MEGA 5 is also a good tool. We have used it in one of our papers, attached to my message. You can find the method we used (how we blasted and etc) for generating the phylogenetic tree in one of the subsections. I have found MEGA 5 pretty straightforward.

thanks @Shahrokh. I have used MEGA as well. But i was looking whether there is any better software or no.

Is there any option in MEGA to remove/collapse tree which is less than 50%. i was looking but can't find that option. anyway, if you know any software which can collapse lowest value or how to do it manually?

Did you make any tree or consensus tree based on distance, maximum likelihood and maximum parsimony...methods? if so, how to do that?

thanks,

and i saw your tree in that paper that you didn't use bootstrap analysis of tree. so does it necessary or if i dont use bootstrap value then how is it?

Note: I am working on fisheries genetics and i saw all phylogenetic tree use bootstrap values. so for me if i dont want to keep less than 50% value then need to remove from tree. if there is any software then would be nice!

thanks

Hey so MEGA is used for very preliminary analysis of your data, just to know how your data is behaving. The trees constructed by MEGA cannot be used for publication, many journals will not accept it, though there are some exceptions, for that you will have to use one of the programs I mentioned above.

Also a bootstrap support is a must, it gives confidence to your result.

Another thing about removing bootstrap values less than 50%, I don't think there is any software for that (although there might be one, of which I am not aware of) but we generally do it manually, it doesn't take much time.

I agree with Arekar. The MEGA 5 is probably a bit preliminary, but we used it for that paper, since they were not picky on that. But I would say that it is really easy to use.

The Bootstrap values are supposed to give you confidence values. It might be good to show them. For example, the value 91 on an edge shows that in 91% of replications (sampling with replacement) and constructing the tree based on whatever method you are applying, that specific edge exists. Unfortunately, I am not aware of some tool putting some threshold for trimming.

Finally, if you'd like to get some idea on the whole thing, you may read: http://statweb.stanford.edu/~susan/papers/bootss5.pdf

Ahasan - just a quick reply to your question about consensus trees.

Making a consensus tree is pretty simple really. As long as you have all the trees you wish to use in the consensus as NEWICK tree files, there are numerous programs that will take your trees and your data set and merge them into a consensus using whatever criteria you decide (majority rule, strict consensus, etc).

The PHYLIP package has a simple little program called CONSENSE that will do this for you (it computes strict or majority rule consensus trees but you set the cutoff for majority rule - 50%, 66%, etc).

Since the NEWICK tree format can represent any bifurcating tree, regardless of the algorithm used to make it, yes, you can produce a consensus of trees from different methods.

Thanks all!

Is there any software to do AMOVA analysis or find the Fst value except Arlequin? thanks