Sequence structure models

In some cases, whole parts of the protein are missing from the experimentally determined structure. At times, these omissions reflect flexible parts of the molecule that do not have a well-defined structure (such as loops). At other times, they reflect parts of the molecule (e.g., terminal sequences) that were intentionally removed to facilitate the crystallization process. In both cases, structural models may be used to fill in the gaps. 

Figure 1 The basis of comparative protein structure modeling. Comparative modeling is possible because evolution resulted in families of proteins, such as the flavodoxin family, modeled here, which share both similar sequences and 3D structures. In this illustration, the 3D structure of the flavodoxin sequence from C. crispus (target) can be modeled using other structures in the same family (templates). The tree shows the sequence similarity (percent sequence identity) and structural similarity (the percentage of the atoms that superpose within 3.8 A of each other and the RMS difference between them) among the members of the family.

During this process of designing sequence changes, models were built and assessed to ensure that there were no obvious steric clashes and that the hydrophobic core was well packed. Furthermore, secondary structure prediction was also used to monitor the progress of change and to choose among different possible substitutions. The final sequence (see Table 17.3) contains 28 changes it had 50% identity to B1 and the similarity to Rop had increased from 5.4% identity to 41%. 

Moreover, molecular modeling is one key method of a wide range of computer-assisted methods to analyze and predict relationships between protein sequence, 3D-molecular structure, and biological function (sequence-structure-function relationships). In molecular pharmacology these methods focus predominantly on analysis of interactions between different proteins, and between ligands (hormones, drugs) and proteins as well gaining information at the amino acid and even to atomic level. 

If the sequence of a protein has more than 90% identity to a protein with known experimental 3D-stmcture, then it is an optimal case to build a homologous structural model based on that structural template. The margins of error for the model and for the experimental method are in similar ranges. The different amino acids have to be mutated virtually. The conformations of the new side chains can be derived either from residues of structurally characterized amino acids in a similar spatial environment or from side chain rotamer libraries for each amino acid type which are stored for different structural environments like beta-strands or alpha-helices. 

The margin of error of a final structural model depends on the sequence or fold similarity to the starting structural template. 

Once an electron density map has become available, atoms may be fitted into the map by means of computer graphics to give an initial structural model of the protein. The quality of the electron density map and structural model may be improved through iterative structural refinement but will ultimately be limited by the resolution of the diffraction data. At low resolution, electron density maps have very few detailed features (Fig. 6), and tracing the protein chain can be rather difficult without some knowledge of the protein structure. At better than 3.0 A resolution, amino acid side chains can be recognized with the help of protein sequence information, while at better than 2.5 A resolution solvent molecules can be observed and added to the structural model with some confidence. As the resolution improves to better than 2.0 A resolution, fitting of individual atoms may be possible, and most of the... 

One Holy Grail in protein structure is to develop tools that accurately predict three-dimensional structures of proteins from their primary sequence information [58, 59]. Many of the best tools to date only go part of the way by using known three-dimensional structures from proteins which share similar primary sequence to model the possible structures of new proteins. This technology still has a long way to go, but the potential rewards would be enormous, allowing a genome sequence to be translated into targets for therapeutic intervention in si-lico, in relatively short periods of time. 

Fig. 1 Solid-state NMR structure analysis relies on the 19F-labelled peptides being uniformly embedded in a macroscopically oriented membrane sample, (a) The angle (0) of the 19F-labelled group (e.g. a CF3-moiety) on the peptide backbone (shown here as a cylinder) relative to the static magnetic field is directly reflected in the NMR parameter measured (e.g. DD, see Fig. 2c). (b) The value of the experimental NMR parameter varies along the peptide sequence with a periodicity that is characteristic for distinct peptide conformations, (c) From such wave plot the alignment of the peptide with respect to the lipid bilayer normal (n) can then be evaluated in terms of its tilt angle (x) and azimuthal rotation (p). Whole-body wobbling can be described by an order parameter, S rtlo. (d) The combined data from several individual 19F-labelled peptide analogues thus yields a 3D structural model of the peptide and how it is oriented in the lipid bilayer...

Kajava, A. V. (2001). Review Proteins with repeated sequence—structural prediction and modeling./. Struct. Biol. 134, 132-144. 

In this chapter, we present several examples of structural models for amyloid fibrils, which we group into general classes. None of these general model classes can completely explain the common properties of amyloid and amyloid-like fibrils however, the Gain-of-Interaction models with a cross-/ spine seem most consistent with what is known. These models combine the structural aspect of the cross-/ spine with the specificity of sequence-dependent interactions to explain the observed diffraction, stability, and self-only association of amyloid fibrils. It is also possible... 

The wide structural application of dipolar couplings is demonstrated by its use to validate models built by sequence homology methods. Additionally, dipolar couplings have been shown to reduce the RMSD between these models and the target structure. One example is the work reported by Chou et al., in which the RMSD of sequence homology models of the protein calmodulin, built from the structure of recoverin and parvalbumin, is reduced using heteronuclear dipolar couplings [110]. 

There are two basic considerations when attempting SDM. One is to determine the amino acids that should be mutated and the other is to decide what to replace them with. The first question is, of course, dependant upon information gathered from previous experimentation in order to target residues that are appropriate. Such information may be derived from biochemical techniques. For instance, in our binding site studies, we have specifically mutated amino acids that had previously shown to be covalently labeled by photoactive ligands. Additionally, we have used comparisons between the sequences of different receptor subunits that correlate with receptor function to identify domains of interest. Chimeragenesis, the technique described in the first half of this chapter, can provide important information in this regard. Obviously, those proteins for which a detailed structural model is available will lend themselves to more rational substitutions. 

---