The problem sounds straightforward for me - the outgroup to root the tree with your samples must be one of the most closely related organisms to your dataset. First try all sequences available and then choose such one. But you also should do many more tests after to get reliable tree (check saturation, calculate substitution model and partition, try different types of phylogenetic analysis - Bayesian inference, Maximum Likelyhood etc.), just doing NJ tree is not the best way to assess phylogeny.

The mcrA gene seems to be popular for classifying methanogens (http://www.ncbi.nlm.nih.gov/nuccore/AY883175.1 ) for example.  But any single gene like that (typical sequenced length is less than 1,000 bases) may not be quite enough for a truly solid phylogeny.  As Oleksandr said, it depends on how saturated with mutations the gene has become across your data set, and factors such as the variation in DNA base composition across the set (AT : GC ratio) and the possibility of horizontal gene transfers, duplications, etc.

If you were studying alphaproteobacteria you might have good luck.  But the major trouble with archaebacteria is that most have not been classified by other methods, so the database is full of mcrA genes from "uncultured/unclassified" environmental samples.  So, now you know your "unknown" is 95% identical to another "unknown" it does not necessarily answer your question about what your "unknown" is.  But sometimes it is interesting to see that your "unknown" from inland mud is 99% identical to a bacteria sampled in the Arctic ocean sea floor mud.

Hi,

First of all, you have to decide to work on nucleic acids sequences OR on translated CDS (mcrA or McrA). The second solution is much better (20 possibilities for each position, .... however, less positions). You will need to translate each DNA sequence according to the right ORF, align them (Clustal/Muscle are fine, but check by hand if there is no problem and refine the alignment). You can add reference sequences, corresponding to known organims of the methanogenic orders (retrieved for example from genbank, but have a look to  RIM DB: PeerJ. 2014 Aug 5;2:e494. doi: 10.7717/peerj.494). Next, you can use several means/parameters to construct the phylogeny (as mentionned by Oleksandr). As an example, see "Phylogenomic data support a seventh order of methylotrophic methanogens and provide insights into the evolution of methanogenesis", Borrel et al, Genome Biol Evol,  2013.

Importantly, check your sequences to detect possible chimeras....

good luck

Thank you for the information. Still, I guess I should narrow my question down a little bit more. Since our lab main focus is in biogeochemistry, so we're just interested in which taxa presented in our research site at the moment. Besides Blasting the sequences and look up their closest known relatives, how do you determine which taxa do they belong to? From my very limited understanding, by constructing an appropriate phylogenetic tree, we can assume that the unknown sequences are at least members of the order of the closest known methanogen in the tree? Are there any other methods for this type of question? Sorry I'm new to genetics.

You are right when saying "by constructing an appropriate phylogenetic tree, we can assume that the unknown sequences are at least members of the order of the closest known methanogen in the tree" (see Luton et al, for this, perhaps 2002, I don't remember well). But check if there are not chimeras and just add some sequences of a few members of each of the methanogenic orders in your alignment. I sincerely think that, even for such a "draft" work, it would be better to work with protein sequences.

Then, perhaps the website www.phylogeny.fr will do the job (just put your fasta sequences).

When you are doing tree just to classify your unknown sequences to the nearest known group still you tree should be reliable and I would invest time to do it properly. Sometimes the position of taxon on a "fast and dirty" and "final" tree may be very different.