gravatar for Mensur Dlakic

3 hours ago by


Various databases of protein families are redundant to a good degree, but not completely. Otherwise, what would be the point of having so many databases if all of them were reporting identical results? This is to say that it is normal here and there to have a domain identified using models from one of the databases but not the others. Without knowing the details of your study or your goals, I would not exclude this protein jusb because it is not identified by IPR004827.

It is for practical purposes impossible to have a single model (in your case IPR004827) that will identify all proteins belonging to a diverse protein (super)family such as bZIP proteins. They are not easy to identify because the zipper part has a strong signal only once in 7 residues, and the rest of it is fairly degenerate. Add a basic DNA-binding region that can also be different (e.g., between AP-1 and Maf proteins), and that makes it very difficult to identify with a single model.

My advice to you is to use multiple bZIP models from different databases, as each of them will be better at identifying a subset of bZIP proteins. As a separate check, you can validate the presence of coiled coils in these proteins using a variety of resources, some of which I list below.

Source link