Biomedical Engineering Reference
In-Depth Information
Modeling by satisfaction of spatial constraints is a technique that uses data from a variety of sources
to constrain the physical configuration of the target protein molecule. Such techniques use data
employed in ab initio modeling, such as data on bond lengths, bond angles, dihedral angles, and the
free energy associated with various molecular geometries. Data from experimental studies, such as
NMR and electron microscopy, can also be used to further constrain the allowable structure of the
target protein. One of the advantages of constraint-based modeling is that the method is inherently
extensible and capable of incorporating any new data or technology into the body of constraints.
Evaluation
In evaluating comparative modeling results, it's important to remember that heuristic modeling, like
ab initio methods, is an experimental process, and that even the best methods rarely achieve
accuracies approaching 70 percent. As such, the modeling-evaluation process is typically repeated
dozens of times before a reasonable target structure is constructed. In this evaluation process,
visualization is key as a first-pass screening tool used to validate gross measures, such as whether
the model has the correct fold.
For a more quantitative evaluation, a measure of target-template similarity can be used. The greater
the similarity of the model with the closest template, as measured by RMSD, the more likely the
model is an accurate prediction of the actual structure. A target model that is radically different from
any of the templates used to construct it isn't likely to be a valid structure. Comparative modeling
isn't intended to model novel structures—an area where ab initio methods shine—but to build upon
existing structure models. Returning to the archeology analogy, if the bones of a newly discovered
early primate are found, the feet should be below the pelvis, and the arms attached to the shoulder
joints, for example. A reconstruction in which the head faces backwards— like the model of a new
protein—isn't likely to be a valid structure. It
could
be a new species with a backward-facing head,
but a close examination of the cervical and thoracic vertebrae (reasoning from first principles or ab
initio modeling) would likely suggest the head faces forward.
In comparative modeling, this inclusion of ab initio modeling to check the validity of bond lengths,
bond angles, and torsion angles is termed finishing. An evaluation of free energy is sometimes used
in the evaluation phase as well. The working assumption is that a correct protein structure has less
free energy (is more stable) than a protein structure that is in an incorrect conformation.
