3.8 years ago by
A multiple sequence alignment algorithm takes a series of parameters that will contribute to an alignment score, i.e., a way of choosing the best alignment as the algorithm proceeds. One way to evaluate how good the alignment is would be to change some of them, like the mismatch penalty or the gap opening/extension penalty, and see whether or not your results change in a meaningful way.
What you get out is, literally, an alignment - an inference of homology - where characters are aligned with putatively homologous characters. Most analyses require that characters be identified as such, especially in phylogenetics and molecular evolution. If you know one of your sequences are ancestral, you can make inferences about derived character states. Imagine you have two aligned sequences:
If you know that the 'AAAAA' sequence is ancestral, then you can say that there are four derived Gs and a deletion. If you have regions that are difficult to align, it might make sense to be conservative remove or mask them (see Gblocks).
There are several applications to visualize your alignments: Geneious, Se-Al, Jalview, etc.
I prefer to use Muscle and Mafft, depending on my needs. There are papers comparing the efficacy of different alignment algorithms you can look at.