 Dear students, as you already know that towards predicting the structure of proteins, there are three strategies. The first one is the homology modeling, the second one is the fold recognition or threading and the third one is ab initio modeling. We were trying to understand the homology modeling pipeline and as you now know there are seven important steps within the homology modeling pipeline. Let's take a look at them to review. The first one is the template recognition and initial alignment. The second is alignment correction. The third is the backbone generation that is the amino acid backbone. The third is the insertion of the loop and optimizing that. The next is the side chain modeling and looking at all the possible confirmations in which the side chains can be inserted on the backbone. This is followed by model optimization and then validation by looking at the known structures. So these were the overall seven steps for homology modeling. So in this module we are going to look at the first one in detail that is the template recognition and initial alignment. If you would remember this was the flow chart for looking at homology modeling as well as ab initio and fold recognition. So we are going to start from here and move towards identification of suitable 3D templates. As you can see so we started from the amino acid sequence of target structure and we now wanted to identify suitable 3D templates which could be used for the prediction of the structure. Now if you had more than one template that you found then of course you have to perform alignments multiple times that is the multiple sequence alignment while if you found one template all you have to do is align the template and the target sequence. Essentially the first step of homology modeling encompasses these four steps. So let's start. So the first step is to compare the sequence of the unknown protein. So the unknown protein is the protein which has the unknown structure. So you compare the sequence of that protein whose structure is unknown with all the sequences of proteins whose structures are known. This of course is found in the protein data bank the PDB database and you can go there and find which proteins have the structures reported for them. Next you compare the sequence of the protein which has an unknown structure against all the proteins whose structures are known. So you blast them and then you find out the proteins which have the highest score. So the highest score will obviously mean that the sequences are similar. You also know that the blast algorithm uses a residue exchange scoring matrix, blossom and pan and that the residues which are easily exchanged during the process of evolution have a low score while the residues that have different properties and are difficult to be substituted they obtain a smaller score. Now function specific and conserved residues they obviously obtain the highest score. So the highest score belongs to the functionally conserved residues followed by the residues that are easily exchanged and the lowest score goes to the amino acids which have totally different properties. So the blast algorithm gives you an overall score for the protein sequence comparison by looking at these of amino acids. Next the output from blast will be a list of proteins from the PDB. Of course the structure is known as well as the sequence and then you have to see which of these proteins are most similar to the protein for which you are trying to predict the structure. So for that you need to improve the alignment and the blast algorithm uses a residue exchange matrix and inserts penalties for gaps and mismatches. Now the target sequence is sent to the blast server which searches the PDB and the list of the templates and their alignments that is the best hits are chosen and they may not be the best one that is if you have blast reporting let's say five different sequences it is not essential that the one on the top is the best one because the structure may have slight bit of variation as compared to the sequence. So in conclusion the template is selected the targets are identified and therefore you can perform the initial step in homology modeling. Next we perform the fine-tuning of this alignment and try to see which corrections are necessary to optimize the prediction of the structure.