Homology modeling .Discuss (25)
The advent of high throughput technologies such as next generation sequencing has led to generation of a lot of biological data which include protein sequences data. The full understanding of the biological roles of protein requires the knowledge of their structures. Experimental protein structure prediction methods consisting of x-ray crystallography and NMR spectroscopy are time consuming leaving a gap between generation of sequences and structure prediction. Computational approaches can be used to develop protein structure models which can be used for rational design of biochemical experiments which include site directed mutagenesis, protein stability and functional analysis of proteins. There are three computational approaches to three dimensional structure prediction namely homology modeling, threading and ab initio prediction (Xong, 2006).
Homology modeling (comparative modeling) is a computational protein structure modeling technique used to build three dimensional (3D) models of proteins of unknown structure ( the target) on the basis of a sequence similarity to proteins of known structure (the template). Two conditions must be met to build a useful model, the similarity between the target sequence and template must be detectable and a substantially correct alignment between the target and the template should be calculated. Homology modelling is possible because small changes in protein sequence result in small changes in its 3D structures. The 3D structures of proteins in a family are more conserved than their sequences, therefore if similarity is detectable at sequence level structural similarity is usually assumed (Pala, et al., 2013). Homology modelling is a multistep step sequential process involving the following procedures: Template recognition and initial alignment; alignment correction; backbone generation; loop modeling; side-chain modeling; model optimization and model validation.
Template recognition and alignment
This involves the selection of all protein structures related to the target sequence and the selection of those to be used as templates. The template selection involves searching the Protein structure databases such as protein data bank (PDB); SCOP; CATH and DALI for homologous proteins with determined structures using heuristic methods FASTA or BLAST and dynamic programming methods sometimes such as SSEARCH for greater sensitivity. Sequences with alignment identity greater than 30% should be selected since they fall in the safe zone and they will be used in the alignment correction to produce a better alignment. Alignments recognize folds in proteins. (Xong, 2006)
Figure 1 Diagram showing zones of protein sequence alignments
Alignment correction
Alignment methods used for template selection usually don’t produce optimal alignment usually for those in the twilight zone thus a more rigorous alignment correction should be carried out on the selected templates to obtain a maximal alignment. This realignment is the most critical step in homology modeling, which directly affects the quality of the final model because incorrect alignment at this stage leads to incorrect designation of homologous residues and therefore to incorrect structural models. Multiple sequence programs such CLUSTALW, T Coffee and Praline are usually used to obtain an optimal alignment to be used in homology modeling (Marti-Renom, et al., 2000).

Backbone Model Building
Once an initial target–template alignment has been built, residues in the aligned regions of the target protein can assume a similar structure as the template proteins, meaning that the coordinates of the corresponding residues of the template proteins can be simply copied onto the target protein. If the two aligned residues are identical, coordinates of the side chain atoms are copied along with the main chain atoms (Pala, et al., 2013). If the two residues differ, only the backbone atoms can be copied and the side chains can be modeled in the subsequent step

Loop modeling
During sequence alignment insertions and deletions produce gaps which require loop modelling to close them. Loops often determine functional specificity of a given protein framework since they contribute to active and binding sites. Loop modeling can be done by two methods either database searching or ab initio method .Database method involves finding spare parts from protein structures in databases that fit into the two stem target region of the target protein . The ab intio method generates many random loops and searches for one that does not clash with the nearby side chains and also allow reasonable low energy. If loops are short 3-5 residues a reliable model can be built but bigger loops present a difficult challenge. FREAD; PETRA and CODA are examples of webservers that can be used to model loops.

(Xong, 2006)
Figure 2 Schematic of loop modeling by fitting a loop structure onto the endpoints of existing stem structures represented by cylinders.

Side chain refinement

As with loops side chains conformations are predicted from similar structures and from steric and energetic considerations. A side chain can be built by searching every possible conformation at every torsion angle of the side chain to select the one that has the lowest interaction energy with neighboring atoms. However, this approach is computationally prohibitive in most cases. In fact, most current side chain prediction programs use the concept of rotamers, which are favored side chain torsion angles extracted from known protein crystal structures (Xong, 2006). SCWRL is an example of side chain modelling program which can be used in homology modeling.

Model optimization

Loop and side chain modeling steps, potential energy calculations are always applied to improve the structure of the model. These steps often cause unwanted structural irregularities such unfavorable bond angles, bond lengths or close atomic contacts. Two methods can be used to refine the structure these include energy minimization procedure or molecular dynamic simulation. Energy minimization reduces the total energy such that the overall structure has the lowest possible potential energy which relieves steric collisions and strains. Molecular dynamic simulation involves process were a protein molecule is “heated” or “cooled” to simulate the uphill and downhill molecular motions. Thus, it helps overcome energy hurdles that are inaccessible to energy minimization. GROMOS is an example of a program which is capable of carrying out both energy minimization and molecular dynamic simulation.
Model Evaluation
The quality of the predicted model determines the information that can be extracted from it. Thus, estimating the accuracy of 3D protein models is essential for interpreting them. The model can be evaluated as a whole as well as in individual regions. There are many model evaluation programs and servers for evaluation including WHAT IF; ANOLEA and VERIFY 3D. The first step in model evaluation is to assess if the model has the correct fold. A model will have the correct fold if the correct template is picked and if that template is aligned at least approximately correctly with the target sequence (Si, et al., 2015). Once the fold has been assessed the more detailed evaluation can be obtained by comparing with the template protein.
References

Marti-Renom, M. et al., 2000. Comparative protein structure modelling of genes and genomes. Annu-Rev Biophy Biomol Struct, 29(1), pp. 291-325.
Pala, D., Lodola, A., Bedini, A. & Spadoni, G., 2013. Homology Models of Melatonin Receptors: Challenges and Recent advances. Int J.Mol , 10(3), pp. 8093-8121.
Si, J., Zha, R. & Wu, R., 2015. An overview of the prediction of protein DNA binding sites. Int J Mol sci, 16(1), pp. 5194-5215.
Xong, J., 2006. Essential BioInformatics. 1st ed. New York: Cambridge University Press.