 Dear students, now we will be moving towards four recognition or threading. Earlier we had looked at homology modeling and how we can predict the structure of proteins whose sequence is known by looking at the structures of the proteins whose structure is known. To give you the background, there were several steps in homology modeling pipeline which included the template recognition, the alignment correction, backbone generation, loop modeling, side chain modeling, model optimization and validation. So all of these steps were performed in a sequence and Modeler is a tool that you can use to perform all of these steps in order towards predicting the structure of a protein if its sequence is known. Now there can be a case where you have the DNA sequence with you and you translated it to get the amino acid primary sequence and then this can be done by using mass spectrometry of course and then so you have to search for the homologs and construct an alignment. This was the condition that is necessary for homology modeling that we were just discussing. Now what if homology is low between the sequence that you are trying to find the structure for and the sequences in the database. So that is the question that we are going to address in this module. So you go and get the sequences from the database and then you perform a blast. If you had homologs with 3D structure known then you did homology modeling. So if the structures were not known or not available then you would not be able to do this. So this is the precise condition in which we go for full recognition or threading. So in this what we do is we recognize the motifs that is the combination of secondary structures. This is a motif and you search the database of secondary structures. This will help you to perform fold assignment. So we will define fold and discuss it in this module in a little while. So why should we use fold recognition or threading? So if you remember this graph we had two portions. So one was above this line. So this is the portion where homology modeling is very helpful in predicting the structure. But if your sequence and identity, if your alignment and identity fall in this range then you go for fold recognition or threading. So earlier we were talking about the red region for homology modeling and now we are going to talk about this portion. Okay so let's introduce ourselves to fold. So a fold is simply a combination of secondary structures that exist for proteins. So the secondary structural elements here they include alpha helices, beta sheets, loops, coils, etc. So a fold is simply a combination of these secondary structures which comes about as a unit. So the common folds for instance are the four helix bundle which are the most abundant and the timbrel. So you will be surprised to know that there are only 5000, about 5000 stable folds in nature. So if I say that if you look at all the proteins whose structure is known you will only be able to find 5000 plus or minus different kinds of folds that are building up those proteins. So it is very useful if we have a database of all of these folds with us which we can use to check for finding folds in protein sequences whose structure is unknown. So fold recognition then becomes a problem in which we find the best fit of a sequence to a set of candidate folds. So here you have several folds in a database and here is your sequence. So you try to find how this sequence matches the folds in the database in a proper way. So the fold which matches the sequence in the best way is selected. So in conclusion fold prediction or threading is simply a protein structure prediction strategy in which we try to compare a sequence with the sequence of different folds. So if the two sequences are close, if they are homologous then we try to see if the sequence can also form that fold. And by creating a combination of those folds we can arrive at the tertiary structure of the protein.