Hello, I have a question for you guys. The perl script rfam_scan.pl uses the program cmsearch to look for a CM model. Wouldn't be better to use cmscan instead of cmsearch? According to the Infernal manual, we should use cmscan if we have a set of sequences and we wanna know the corresponding or closest CM. Don't you think so?
Hi there, Rfam project leader here (for verification you can check our Twitter feed; we've linked to this question)
Yes, in future we'll be providing cmscan instead of rfam_scan.pl. Historically, rfam_scan.pl needed to include the BLAST filtering step, which reduced the search space the CM needed to cover, making the problem computationally tractable. However, INFERNAL 1.1 now uses HMM-based filters as part of the cmsearches, meaning the blast filters are redundant.
As far as I recall, cmscan (like hmmscan) and cmsearch (like hmmsearch) return the exact same results. The big difference is that the former can't be parallelized without special effort.