A homology model is built by determining a suitable template followed by aligning the sequence of interest and from this a model is created. High-quality models can be created for sequences showing a high sequence identity with the template. For an identity lower than 30% the alignment already might not be good enough to build a good model.
The nuclear receptor superfamily is of major interest to the pharmaceutical industry.
The X-ray structure of 16 different nuclear receptors are known, but this fast-growing family already contains over 1300 sequences annotated as such only in Swiss-Prot and TrEMBL. Building homology models for these by hand would be a waste of time.
By superposing the known structures and removing structurally non-conserved residues we were able to build a 3D alignment for over 450 sequences, using profiles. The superfamily can be divided into subfamilies which can be aligned using these profiles.
AMB is a program that uses these profiles and family distribution to build models automatically.
If a structure is known in a certain family larger parts of the structure can be modelled. In contrast, if there is no structure known for a certain family and the sequence identity to the closest template is very low, only the structurally conserved parts can be modelled. Because AMB uses only the profiles and the family distribution it can be applied to any set of proteins for which more than one structure is available.
