Hello all,
I have a task of clustering a large compendium of plasmids. They are quite different in size, and also not necessarily rotated the same way (that is, many sequences are identical, but are not starting at the same point).
What I want it to cluster them according to CUMULATIVE sequence homology; i.e. not the best blast hit, but rather the overall coverage of homologous regions on both sequences.
It clearly has to be a local-type algorithm (because of the substantial difference in lengths), but local version of cd-hit-est does not perform best because of the rotation issue (see above). Blastclust did even worse (I didn't dig too deep into why, and it's kind of hard because of very terse output).
Thus, I would appreciate any suggestions. Thank you in advance!
Try
mashdistances? That might give you the resolution you want to cluster them, and AFAIK is ‘reverse complement’ insensitive, as kmers and their reverse complements are considered the same kmer.That might be good enough for a first pass?
Excellent, thank you. Definitely worth a try.
Guess what, in the newest NAR issue there's a paper about a plasmid database that is using mash strategy to cluster them. Talk about timing :)