Parallelize Alignments Using Clustalw-Mpi

Question

[David M]

Are there many people on here who have experience using the parallelized version of ClustalW, ClustalW-MPI (http://www.bii.a-star.edu.sg/achievements/applications/clustalw/index.php)? From the readme, I see that it was released in 2003 and based on a version of ClustalW from 2001 (v1.82). From what I can see, it doesn't look like its being actively supported.

I'd like to perform some large (>10,000 sequences) alignments, and using a parallelized version of ClustalW would make it much faster. Are there any reasons to use/avoid ClustalW-MPI?

Answer 1

[Andreas]

Hi David,

my guess is that the ClustalW-MPI code is not actively developed/supported anymore.

However, there are at least two programs specifically designed to be able to deal with such a large number of sequences and they don't require parallel computation. If your sequences are protein sequences, try Clustal Omega. In case of nucleic acid try Mafft with Parttree.

Andreas

Answer 2

[Neilfws]

I recall trying out ClustalW-MPI when it was first released and seeing some performance improvement on a small cluster.

One issue is that it requires either MPICH or LAM-MPI, which were the two competing MPI implementations at the time. I believe that Open MPI is the "preferred" implementation at present. I still see packages for the older MPIs in my Ubuntu repository, so in principle you should be able to download and compile against these. However, I don't know whether MPICH/LAM-MPI is available on the cluster that you want to use.

Another issue, as stated already, is that the software is unlikely to be supported anymore, given that there have been no updates for 8 years. So if you do have problems, you'll be on your own.

Finally there is almost sure to be a better, more recent solution. And remember that ClustalW, MPI or not, is the alignment solution that "everyone" knows, but not necessarily the best or most appropriate.

By all means give ClustalW-MPI a go, but investigate the more recent alternatives first.