Hi, I have used CLUSTAL W to align an 18S nuclear DNA sequence with it's closest 8 matches in GenBank (all 8 matches were 100% QC, 96 - 98 ID scores, and the same 8 appear in a publication with K2P differences of no more than 3%. However, in an ML tree topology my sequence is placed in a clade (100% ML bootstrap value) with just one of the above mentioned taxa, and the rest of the group are in separate clade subtended by an unusually long branch length. It looks like there could be an alignment issue causing the problem, is there an alternative alignment tool to ClusatW that might help in this instance?
Try the different Multiple Sequence Alignment tools that are available through EBI (http://www.ebi.ac.uk/Tools/msa/). There are brief overviews about each one of these tools and for what they serve. In your case, I would test WebPRANK.
Good Luck
Use Multalin tool available for free online http://multalin.toulouse.inra.fr/multalin/
you can try different strategies and compare the results using mumsa (http://msa.sbc.su.se/cgi-bin/msa.cgi), use guidance (http://guidance.tau.ac.il/) or trimal (http://trimal.cgenomics.org/) to identify reliable regions.
The MCoffee tool is nice in the sense that it uses several other tools, and looks at the consensus. In this way you are less biased by the choice of tool.
Another option is the Infernal software, since you have 18S sequences. I would expect there is an 18S Infernal model that can be used to stabilize the alignment. I have only experience with 16S (prokaryotes).
A lot of progress has been made on alignment strategies in recent years and tree guided alignment or simultaneous alignment is definitely something to consider depending on how computationally challenging the problem is. There are iterative tree guided alignment procedures such as Sate (I think it's called PASTA now) which can handle fairly large problems. Some really interesting work has been done on Bayesian estimators that treat the alignment, topology, branch lengths, and model as parameters (BaliPhy is a good example), but these are limited in number of taxa due to the computational complexity. There are also lots of progressive alignment algorithms utilized in a suite of programs, you just want to use the algorithm appropriate for your data. For example, different algorithms make certain assumptions about the data, such as 1 alignable region with variable flanking regions (L-INS-i implemented through MAFFT) vs. the assumption that all of the data is actually alignable (G-INS-i implemented through MAFFT).
- Badges
- Science method