 Dear students in this module will be moving into the ab initio methods We've already looked at homology modeling as well as fold recognition and the 3d1d Bowie algorithm Why do we need to study ab initio modeling? so the simple reason is That there can be cases where the previous methods they fail So let's get a brief background on ab initio So ab initio the word ab initio means from scratch so in the homology modeling We were using the PDB structures in the fold recognition process We were using the fold databases, but in the case of ab initio the ab initio modeling We don't have any such database on our disposal so ab initio modeling then Utilizes the infants in hypothesis that every protein structure tries to take the confirmation with the minimum energy so using that hypothesis We can say that if we have predicted a structure that has a very low energy Then most probably it will be correct Of course it can be correct, but there can be situations where such a prediction may not be biologically useful So we'll discuss that in this module and the later ones So basically ab initio methodology tries to predict the structures by minimizing the energy of the predicted structure So you can take up any sequence for which you don't have any alignment or template and then try to fold that structure sequence into a structure and Compute the energy of the resultant structure Now several things come into our mind. First, how do we fold the sequence? Second, how do we compute the energy? Third, how do we minimize this energy? So all of this is very important and is a matter of great debate in ab initio methods area As I just mentioned there can be cases where such sequences are folded by ab initio methods and The predicted structures are not very accurate. So of course we need to validate these structures as well Lastly the accuracy and applicability of ab initio methods is limited by our own understanding of thermodynamics and underlying energy Transitions in different folds So therefore our own understanding of these dynamics Limits the output the quality of output from ab initio methods Now let's discuss the limitations of such methods if we have a sequence and we have folded it into a structure Then obviously there can be many confirmations that we can fold the sequence into Upon every structure that is output by folding the sequence we need to compute its energy as well So this means that it is a computationally very expensive process So if the energy computation is Expensive and that you have to do it for so many different structures ab initio methods tend to solve only small sequences and their Structures if you have a large sequence, let's say 100 100 kilo 100,000 kilo Dalton protein, which is a very large protein You may not be able to use ab initio methods because it may take very very long or maybe forever to compute the optimal structure So typically we predict the structures for proteins that have less than 100 residues you can also consider Cleaving your sequence and Predicting the structure for portions of your sequence and then bringing them together using ab initio modeling So as we discussed earlier if you found a matching sequence in The PDB Then you're happy doing homology modeling. So you had your sequence you went to the PDB and you looked at all the homologous sequences and Their structures and then you predicted the structure for your own sequence So let's say if you selected One structure then you scored this match however There was a situation Where you could not find These proteins So if these proteins were not found in the PDB, then you had no choice except to move to Ab initio modeling. So in ab initio modeling What you have to do is you have your same sequence here sequence and you try to fold it into all the possible Structures and Compute their energy So the one with the minimum energy is Selected the one with higher energy is Rejected of course there could be a situation where you find something in the PDB, but It has a low score. So in this case You used for recognition or threading. So this is Why ab initio modeling is quite useful for cases where you don't have The structure from the PDB In conclusion the ab initio methods they rely on computing the energy of structures and Once you compute the energy you select that structure, which has the minimum energy and Once you have calculated the energy that is the minimum for a structure Then you see its plausibility in the biological context as well Because there can be a situation where ab initio methods can predict a structure that does not have a biological meaning