 These students in this module will discuss a very interesting topic. In fact, a very interesting pitfall that you may encounter while doing ab initio modeling. So the title of this module is global versus local minima. To remind you of the background, the protein structure prediction using ab initio methods, it relies on the bonding or unbonded state of atoms within a protein structure. Now there can be a case where you have folded a protein using ab initio methods and you have computed its energy to be the lowest. However, there may be some other protein structure which may have even lower energy. What I mean to say is that there may still be further possibilities for optimizing your protein structure. Now if you are doing ab initio modeling, you would love to have a structure that has the absolute minimum energy. But it's not easy to obtain that structure because such remote and really minimum energy structures are difficult to find. So let's discuss in this module how we can find such structures. So here you have a sample protein which has two alpha helices and several beta sheets and in each one of these secondary structures you have these amino acids of course which have been labeled for a certain portion of the structure and each one of these amino acids has numerous number of atoms. Now these atoms they will be interacting with each other and there may be a chance where these atoms cannot interact with each other. Now if you want to compute the overall structure's energy then you have to add all such situations and arrive at a overall energy value. So this is done by using force fields and here is your structure that you have placed inside the force field and this structure is surrounded by water molecules and you are just counting how many bonds are made. The overall energy again is computed using the same relationship. So now the structure that you have here has in fact evolved from an alpha helix let's say and then it took this shape and then it took this shape and in the fourth hydration it took this shape. So an alpha helix after several evolutionary steps has attained this structure and this one has the lowest energy. So here in this slide there is this curve in which we have these valleys and these valleys each one of them it represents an energy value. Just to give you an example so if you have a ball and you set it into motion so most probably it will fit here or fall here and then continue falling until it reaches here. So once the ball achieves this position then it cannot move any further. So this analogy can be extended to the protein structure energies wherein the proteins initially have this energy value so obviously in this direction the energy is and here energy is high. So this is a protein with a high energy content due to a lot of non bonded atoms and as you can see the structure is evolving and the energy is becoming smaller. So by the time the structure arrives here it has taken the shape of a beta sheet instead of this very large blob of amino acids and you can say the structure is the final structure. So you have to consider the structure's energy but a case can happen wherein your energy landscape is like this. So now if you consider the same example of a ball that is moving in this direction then obviously the ball will settle here and fall here and fall here again but after this it will be very difficult for the ball to actually go and move in this direction. So even though if this ball were to fall here and then move here then it will become impossible for it to go and shift its energy like that. So there can be this uneven energy landscape and your protein structure that is predicted may just be one of these positions. So these are the local minima and you need to perform global energy optimization. So you need to know which positions are available in terms of their lower energy and the protein can take one of them. So in this case you can easily see that this protein structure has a higher energy and this protein structure has a higher energy as compared to this one. So this is your good prediction as compared to these two. And then these proteins as I just mentioned they can fold into this structure. So the best case function can be a clear energy function which can minimize the energy of the structure that you are trying to compute in one step but of course there can be situations where you have an uneven energy function and that can create problems for you because you are not sure which minima is the global minima. So you need to have a viable path as I just mentioned in the example towards the local minima. So the optimal energy function then is in real life we cannot have a global function for minimizing the energy. There can be many many possibilities and it is difficult to explore all of them. Hence instead of going for a global minimum function we find the optimal energy minimization function. So it is easier to design and compute and therefore you can play with that function more easily. Also as a result the native structures that are reported from the prediction are not necessarily the global minimum structure. And there is no clear way of choosing between such functions.