I need to study domain gains/losses in species of protists that are quite divergent from each other, I want to align proteins based on domains and visualize the domain pattern on the multiple alignment to infer gains and losses.
Hi Nicola. I do use Jalview and find it very useful for domain study (biochemistry and 2d prediction based on putative homology of similar sequences). You can even have an idea of the 3d (function) of a given nuc. sequence (if 3d data model use available for your protein) and visualize the position of an amino acid on the 3d projection of similar sequence. I use it for cloning domains with specific putative functions and also for studying the evolution of a protein. Hope you will find it useful!
I've used the program iprscan (interproscan, http://www.ebi.ac.uk/interpro/search/sequence-search) to identify protein domains from transcriptome or other types of gene space assemblies. There is an executable for iprscan. For smaller scale data sets (a handful of sequences) I've used NCBI's conserved domain finder, http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi. It has several databases, including interpro, that it searches from. As for studying gain/losses you will need to put your observed patterns of domains into a phylogenetic framework with an appropriate outgroup. This would be the "big picture" of domain evolution at a species level. There are ways of doing this and being statistically and phylogentically savvy about it; have a look at the R package phytools. If a domain is conserved and is found in multiple sequences you can try to build a tree from just the domains, which can be extracted from your sequences given the coordinates from a domain finder. However if the domain is small it may be hard to recover strong support at nodes of the phylogeny. I hope some of this helps!
You might have to do the work in two steps. First, by applying PSI PRED and Dom Pred, identify the domain boundaries then by applying multiple alignment tool, compare the structures.