Mike White wrote: > I am modeling a homopentamer using three structures of homologous > homopentamers as the template.The monomers are 210 aa's. Following the > advanced tutorial, I first aligned the three known structures using > salign with the three cycles of refinement, and got good alignment. In > the tutorial, the next step was to then align the test sequence with the > aligned templates using align_2d. After approximately 3900 seconds this > produced what looks like a nice alignment, with structural features such > as cys involved in disulfides and conserved aromatic residues lining up > between all four sequences (three templates, one test). I then did the > alignment using salign instead of align_2d, following the description in > the online manual for salign (examples/salign.saling_align_2d). Once > again, a nice (but slightly different in the regions without much > homology, but still with the other features retained) is obtained. The > surprising thing is that this alignment only took 17 seconds, a 200-fold > improvement in time). > Both alignments gave acceptable-looking models. > > Are the two algorithms so different that salign is that much faster? Is > there any advantage to using align_2d instead of salign?
No; salign is a superset of the other alignment routines, but uses the same underlying code (the advantage of salign is that you can combine information from different sources to build your matrix). It is significantly faster in your case because the salign example sets max_gap_length to 20 (rather than the default, which is rather larger). This dramatically reduces the space which must be searched.
Ben Webb, Modeller Caretaker