Making an approximate 3D model of an enzyme

Sometimes we want to check out how an amino acid sequence from a database might translate into a 3D dimensional structure. This is straightforward enough if there exists an x-ray crystal structure for that exact enzyme in the Protein Data Bank or, like often happens in the nitrile hydratase class, there is very limited variation within sequences so overlaying a short portion of the sequence "by eye" on an existing enzyme is possible.
We are currently working on nitrilases, and despite increasing interest in their use as biocatalysts, the number of nitrilases (i.e. enzymes that convert nitriles to carboxylic acids not the other looser biochemical definition) which exist as structures in the PDB s precisely one. It is from Pyrococcus abyssi and is pretty much restricted in substrate tolerance to small aliphatic nitriles like fumaronitrile. It should not be a surprise that it isnt a great model for many other nitrilases.
If the PDB has come up short we tend to use a link out of the Uniprot database to give a prediction of 3D structure. So if you want to see an estimate of what the nitrilase from Aurantimonas manganoxydans looks like, than the link labelled "ModBase" under the subtitle "3D structure databases" leads you to a page of predictions and further tools. For this specific enzyme it is suggested that a mouse nitrilase superfamily structure contains the best 3D match.