 Dear students, in this module we'll move towards the last step in the homology modeling pipeline. So now we will talk about the model validation. Till this point we have constructed a structure for the protein by using the protein sequence and the protein structures from the PDB. We have created its backbone, we have inserted the rotomers, we have also put the gaps and replaced them with loops and coils. Moreover, once we have done that we have then optimized the structure by putting it in MD simulation and then we have selected the structure with the minimum energy. Now having done all of that we have still one problem. That is we have to check each and every torsion angle using the Ramachandran plot of course as well as bond lengths, the bumps in the structure. So there may be some exceptional protrusions from the structure that should not be there and this is what forms the crux of the model validation. So let's start. So as I just mentioned we now have a protein structure that we have predicted and that it has the minimum energy and therefore is very stable. This was obtained by using molecular dynamic simulation. After we have predicted the structure we need to evaluate the structure for bumps, bond angles, torsion angles, lengths of the bonds that are there and several other properties such as distribution of polar and apolar residues. So for instance if you have a protein like this so this does not have a obvious bump however if you have a prediction such that you have helices and then there is a portion in the protein that is protruding out of the main structure so you want to reevaluate this structure. Also the bond angles in the backbone should follow the Ramachandran plot. So this is called the Ramachandran analysis and it is an integral part of the protein structure prediction problems. Similarly in the next step you have to see the angles of the rotomers and see if these angles are valid. So all of these angles need to be within a certain range for each amino acid. So remember that each amino acid has a certain preference for the torsion angle that is the angle with the side chain. Next you have to really see what are the bond lengths. So if the structure that you have predicted has these bond lengths that are very long or very short then you would want to adjust the structure such that the bond lengths they stay in the tolerable range. For instance hydrogen bond should have a length of about 2 to 3 angstroms. The other covalent bonds should be according to the atoms involved. And lastly other properties such as the distribution of polar and non-polar residues should be normal. So in the structure if you have a polar or a charged residue inside the protein structure then your structure can have an unstable core and therefore may be an incorrect prediction. So the important points to remember is that we can tolerate the errors, slight errors but such errors should not be in the active sites of the protein. So the active site is that site of the protein that is responsible for the function of that protein. So the hydrophobic core may have slight variations but the active site of the protein must not have a lot of variation. So now that we have performed the entire 7 steps of homology modeling we must also critically analyze what are the shortcomings of such an exercise. Because the homology modeling is based on the template structure so it means that anything that you predict by using the template structure is only limited to the structural design of the template and therefore there is a large bias towards the protein structure. Also if there are conformational changes in the protein whose structure is unknown that is the target then because you are using the template you cannot predict those conformational changes and therefore you can miss those conformational changes. And lastly the homology modeling pipeline cannot help you to discover newer sites for binding of drugs and other such novel behavior. The fundamental reason for that is because you are using the knowledge that you have from the structure of the known protein and therefore you are limiting yourself to the structural prediction for such a protein only. Now how do we overcome these limitations that are set in place because we are using a template structure. So there are two other ways for homology for the prediction of the protein structures as well. So one of them is called threading and the second one is ab initio modeling. We will look at these approaches later but the point here is what to do if homology modeling fails and therefore you also have other methods to predict the protein structure instead of homology modeling. So we will examine the online tools that are available for each one of these strategies later as well.