trRosetta subpage

In Pfam, apart from the structures available from PDB, which are resolved experimentally, and the AlphaFold structures predictions, you can find a second structure prediction derived from trRosetta models, which are generated using Pfam domain predictions.

Let’s have a look at an example, the putative sugar diacid recognition domain (PF05651). This is a bacterial domain that may be involved in sugar recognition, found in proteins characterised as carbohydrate diacid regulators.

The trRosetta model can be accessed from the trRosetta Structure subpage, on the left-hand side menu (hard refresh the page if it doesn’t displayed immediately). PF05651 has and alpha-beta fold with a long alpha-helix at the C-terminal (Figure 16, remember hovering over the structure to see the amino acids and their position).

If we look at the AlphaFold prediction available in this entry (Figure 17), we see the full length protein structure prediction comprises three domains. This is of great help in this case, as we don’t have full structures experimentally determined. The domains are highlighted using the same colours as the ones in the domain organisation subpage (the putative sugar recognition diacid domain of PF05651 in green, the diguanylate cyclase-like domain in red and the DNA-binding (helix-turn-helix domain) in blue).

If we compare both predictions, we can see that they agree: both predict that the N-terminal domain has an alpha-beta structure followed by an alpha-helix. Moreover, the other two predicted domains are also in agreement with the information available. However, these predictions must be interpreted with the confidence scale provided as some regions are more difficult to model.

Questions