 Okay, so I'm trying to guide you a little bit, demonstrate some of what you can do with Hadock. I intend to show you where to find information and when to find all kind of tutorials that you can run on your own. Let me start sharing my screen. Okay, answer to the question that was asked during the talk comparison with over docking software. So you see here, flex six, black, surf light, blue, gold, dark blue, glides orange and the shape and pharmacophore results. So shape will be green. Remember that for the flex, surf, flex, gold and glides are bond docking results. So they are national changes. You take the crystal structure of the complex, you just separate the molecule and try to dock it again. While for the Hadock results, the shape and the pharmacophore one, we generated conformation of the ligands from smile strings, from scratch and we're using the conformation of the receptor which is not the crystal structure of the complex. But what is the quality of the best model generated irrespective of scoring? In unbond docking, in this scenario, you see that our shape protocol is repetitive with the other ones that are bond. So we're not comparing the same thing. And if you look top one performance, the shape one is just percent, unbond docking, gold and glides are reaching just 5%. So I think that should make clear that if you can find template, this is really a competitive protocol. Okay. Now, I'm going to guide you a bit to the sources that we have at hand and we're going to do a sub portal. So at least I'm going to do a submission. You can follow the steps and you can also look at the results and then take a look at things together. You will find a lot of information on our group page which is bonvalab.org. And here you find, well, I mentioned the software. If you go to software, you will find Hadoq, 2.4 version, official production version. We are working. Hadoq 3, which is a new, complete rewrite of Hadoq, which will be modular so you can cut the protocols. But this is changed. So I will not recommend to use that one, but this is, you can go to the repository. It's fully open for now. 2.4 is our official version. But so you can find also all kinds of interesting software, tools, the use of Hadoq PDB tools for working with PDB files, modifying them, and other tools, data sets. If you are more into optimizing functions or you're interested, so we have a lot of models that we generated by docking for different type of components, which are basically, which can be used if you are into machine learning and things like that. And the link is in the chat, if you are searching, where I'm just going to go to the software page of Hadoq. Just show you, there's an online manual. You can find a lot of tips preparing PDB files for docking, defining restraints, defining all kinds of things that I have shortly. Project setup parameters, what are the meaning of the different options, a lot of parameters that you can change in Hadoq if you know what you are doing. Frequently asked questions, analysis, scoring, what is the scoring function. And important, if I go back just to the, you also have a best practice guide. So if you're going to use Hadoq and you've never used it before, why is it to take a look at this guide, something we wrote under BioXL actually. And so how do you prepare structures for Hadoq? How do you use information? What kind of system are you docking? So you can see small molecules, we also support peptides, nucleotides, protein, grading, so for each of those, there might be some input that you can find there. How to analyze the dock and dock. So really it's recommended that you take the time to look at that if you want to use the software. Now, if you are interested in learning to use the software better, you should visit our education webpage because this is where you're going to find tutorials. They're related to courses that we are giving here in Utrecht. They are also recording of some of our lectures like the BioXL summer school this year where they are two lectures of about 45 minutes. And then we have this tutorial section. Again, the best practice guide is here and we have the tutorial for version two, four. See that tutorial for the previous version of Hadoq which also is associated with that. But the server is going to be discontinued probably at the end of this year. Tutorials for over-related topic being some of the summer school tutorials that we've been doing. So let's take a look at Hadoq 2.4. Now, if you want to install Hadoq locally, you can do that following this installation tutorial which also guides you through different scenarios how to look in. And then we have different tutorials going from rather simple. So this will be the most simple tutorial, basic protein, protein, dock, NMR data. We have a tutorial demonstrating the use of cross-link data from my spectrometry. In this tutorial here, we are talking of an oligomeric puzzle. Actually, the answer to that is not in the tutorial. This is something I'm using in detail where you are trying to base on cross data to figure out if you are dealing with a dimeric, trimeric, tetrameric or pentameric system. We have done work on doing template-based modeling of protein-protein complexes where you can derive C-alpha, C-alpha and it's from homologous complexes and use those modeling. So this is what this tutorial is about. Here you find a tutorial based on the previous Capri target with 70 where you are doing T-body symmetrical docking without any information except for the symmetry of a homo-detramer. Here you find a small molecule binding site. This is against a large receptor. Actually, it's a transmembrane receptor where in first instance, assume we have no knowledge of the binding sites and we do docking, tire surface, do a statistical analysis of the contacts that are made to predict binding sites, then target a specific binding site. So if you remember, I told you that we had different scenarios for small molecule docking. So this will be an illustration of the first, at least the second stage when you do, we set up the docking targeting the identified binding pockets. It's probably not the best way to do small, without dock. The shape-based protocol is way better but this is illustrating that protocol. We have a brand new shape restrain on molecule docking tutorial which is basically illustrating the tutorial, the protocol that we just published and mentioned in my talk. And this is what I want to take you through today, bits of it because this is more complex to run into only one hour. I also show you in my talk, the anti-gen docking. And we have a tutorial to do that. Different type of data, no information on the anti-gen. So you take the entire surface or using information on the anti-gen. So it's more directed docking. And finally, speak about, well, I gave you an example of Membrane Pro who published recently this year a paper where we combine light dock as docking engine, a dock as refinement engine to model membrane processes. And this is illustrating how to do that. So these are the tutorials that we have to point you to courses and the structural bioinformatics and master course that I'm giving in U-Test. But this comes with extensive tutorial which consists of three parts doing homology modeling. So here we use Swiss model in the latest version of the tutorials to run molecular dynamic simulation of peptide using Gromax, where we describe all the setup and all the analysis. And then we set up protein peptide. We don't have a protein peptide tutorial in a list of tutorials, but you find one here. Since some of you are on peptides. And actually, so this tutorial is using to run Gromax, we use virtual machines, VMs, cloud machines that you can access at MMR box. So again, trusted lots of information there and lots of tutorials that you can run at your own, on your site. So for today, what I want to do is to go into this shape restraint protein small model. We're going to do more drug design like so I want to really particular one. So, really you are following me a little bit. So all our tutorials are organized in a similar way that you find on the website. So this one, we're going to do this template-based shape restraint modeling of a protein's ligand, okay? We're going to use the website details of the paper. Well, this is the preprint pointing to bio archive, but the paper is not published. And what is important, all the tutorial of these guidelines. So we use a color coding to define actions that you following the tutorial should do. So orange will be used. So you should think about data and try to answer it. Blue is an instruction. So you need to do something. For example, in putting uploading data to the web portal. Green we use for primal. So for visualization of the molecules, this could also be camera in some tutorials where we use MS or EM data, we use camera for that. And black will be a Linux prompt command. So something that you will not need to type at the Linux level. Of tutorials where you don't need at all to use the Linux, the command line level. So you can do everything. Basically you don't depend on the operating system. So we'll find more and further the web portal. Now for these particular tutorials, if you want basically run it on your own, of all kind that we have, we are providing a tar archive which contains all the data and actually pre-generated data that you can do the submission. So if you want that, you should download this file by clicking on here. And I did that already. This file on my desktop, you see it here. Shape small molecule, okay. You also need for this tutorial, since we're doing a lot of small molecule work, you will need to install where. So we use open Babel tools, we are the kit. So I think it's too much to ask for today during the time of the tutorial, if you want to do it on your own when you have the time. Download the file, now you will have all the data that you need to upload to the server. So we are basically providing data that have been calculated for several steps, okay. You see this tutorial also uses PB2 specifically. Now, if I wanted to really do a submission, a real submission, well, I should register for accessing Hadock. And what we have here is some, what Hadock is doing. You can find the movie view in my presentation about the docking process. So but you are not, you are now all experts seen into the presentation. So what are we doing today? So we want to model the binding of this particular molecule that you see here to a protein. So that's part of, it's one of the complex, you know, due to the data set and it's actually a PDB ID 1D 3G. Protein you see at some co-factor associated with it and we are going to dock this particular one, okay. So this is the binding site. So we have smaller molecule in the receptor than just a link. And the protocol that we're going to follow is the shape restraint protocol. The tutorial also demonstrates the user aquaphore-based protocol, but we won't have time for that today. Other steps basically that we need to follow. So first we identify and download potential templates of interest for target of choice. Again, this protocol only works if you can identify in a PDB some complexes of the receptor of choice, which has some overlegant associated with it. And often for say drug targets, you will find quite a lot of information. If there is nothing in a PDB related to your particular protein, of course, you cannot follow this protocol. So you should go back to the Moabini show, ligand docking one that I show you as one of our tutorial. So once we have identified all the possible templates, we will have to select the best template for the particular ligand that you want to dock, prepare the file for the docking, do the docking and analyze the data. Now, how do we have my templates? We do that by using, by the way, if you read the, there's a link to a Git repository where all the providers, so you can do a lot of this in a more automated fashion. So in this protocol, to identify templates, we need to RCSB the protein and use the advanced search that will allow you to basically retrieve all PDB entries. In this case, we want to use this. So you need to provide a sequence of your targets. So we use a sequence information of one D3G and we want to only retrieve targets that have 100% identity, okay? So that's a, I can demonstrate that, open this link in a new tab, see now I'm RCSB, please a little bit. So I'm interested, I want to provide a sequence and the PDB is one D3G if I'm correct and D3G, yes, okay? And you specify here that you want only to look at all the entries in the PDBs that have 100% identity. Okay, and if you do this, you click on search, search. Yeah, let's see, oh, PDB, yes. Click 100% so now the sequence is uploaded, search. Okay, three trees, now you start getting a list of all the related entries that have different ligands. Okay, so it's possible to download from the PDB file, basically a file that contains all the data. Okay, and of course, if you have downloaded the data from the tutorial, since the PDB is constantly that what you download today will be different than what we are providing here in the tutorial, okay? So you want to generate a port within the ligand preset and save it in CSB format and show you what that looks like. So I'm here, you have all the software related to this tutorial, we are looking at data against CSB that's the file saying, and what you see here is basically a list where you find the PDB end here. A lot of information about the ligands in downstream but there's also the in key key and the different molecular weight formula, all of that. We are interested here in the smiles. So this is basically what the search in a PDB returns. So a lot of different ligands for these proteins that have 100% sequence identity to your target but they have different ligands, okay? And we only want the protein itself contains also these tools, so we want to keep the entries that have those cofactors but also over ligands. And you don't want to, excuse me, you don't want to keep there's also some time crystallization junk that you are getting out of the PDB because when you crystallize structure they might be all kind of small molecule in the buffer that will come and that are part of the crystal structure. So you really want to look at entries that have ligands in the pockets that you are targeting. This in a data set that we are providing for the tutorial this has been provided to you. So this is these filtered file ligands filtered.csv. Now we want to select the template basically in which one of all the ligands in this list is the most similar to the ligands that we have to doc. And for this we're going to calculate the similarity between our target ligand and the ligands that we found in the PDB using the Thresky coefficient which compute basically the maximum substructure and we use the RDP implementation for that. So we provide, so basically get the small strings from this list that you downloaded and you only output basically the PDB ID and then the information about us you're going to get this file. So this file again is provided in the data so I don't need to run this command. I'm just going to show you what is the contents. So you see here now you have the smile strings of all those ligands which is a one-dimensional strip describing the chemistry of the ligands and then you have a name of the PDB file where you find these ligands and the name of the ligands in that PDB entry is B5O that you see here to the name of the ligand in that particular entry. So that's what we are doing. So now we have us calculate what is the basically the maximum common substructure. So in the tutorial we have a screen to give you that and then you could add basically the Thresky metric, okay? And this is going to give you the similarity of the ligands to your target ligands. So let me just show you what's in there. Similarities. So now you see we have calculated the similarity of the different ligands. So you have here the PDB ID of the tag B you have the name of the ligand in that particular PDB and the similarity to your target ligand. So if it's exactly the same, the value will be one and they are as far as sorted. So you see here that there are two entries that are very high similarity to the ligand that we want to dock, they are not exactly the same but they are highly similar. And the higher the similarity, the better the docking results are going to be. So this is also demonstrated in a paper. So you see actually you have about six templates for which the similarity is 0.7 and then it drops to 0.5 and lower. And some of these are really very dissimilar point impact. So I'm good. So if you want, so the top two, so you see here the similarity, the top two ligands. So there is again two pH and seven K2U. So they are very similar at 956.942, very similar. So then you have to also, if you're going to, you have to make a choice. So what is important here, it's not black and white of course, check the quality of the structure itself. So if you compare the structure, the resolution of the second one, which is slightly less similar is 1.72. And the other one is 2.4. So that's one good argument to select one for the other. Further, the top rank structure is also in some loops in some region of the structure, which might, here there is a break while here the loop is going around here. So this is also going to basically, if you do further molecular dynamics with no structure, you probably want the structure which has the highest similarity, okay? So what is high similarity? So again, in the, so if I look now at the seven two, I'm going to load the one D3G. So this is what we are targeting, what we want to reproduce basically. And this is not the template that we would select based is that we have done now to align. So now you see alignment of the two molecules. These are the cofactors, okay? This is crystallization buffer. These groups that are part of the crystallization buffer. So it's all to do modeling that you should look at what is in a crystal structure or the PDB file that you are using. These are the crosses here, the ligand that we are interested in and you see close to that there's also some over-molecularized. So this is what we want to generate basically and this is what we select. The similarity in terms of 3.946, so it's very similar. There are some differences in this region here and they are in these aromatic rings, but otherwise it's quite similar. So it should be quite obvious that if we use this ligand as a shape to guide the dock, it should get very good results. So basically, it's like when you do protein prediction, you can find an homologous structure, you're going to use that a molature rather than do an abinitio prediction of your protein. Now we do the same, but now for small molecules, the ligand that we are interested in and you see close to that there's also some over-molecularized. So this is what we want to generate basically and this is what we select. And the similarity in terms of 3.946, so it's very similar. There are some differences in this region here and there are some in these aromatic rings, but otherwise it's quite similar. So it should be quite obvious that if we use this ligand as a shape to guide the dock, it should get very good results. So basically it's like when you do protein prediction, you can find an homologous structure, you're going to use that a molature rather than do an abinitio prediction of your protein. Now we do the same, but now for small molecules. So we have identified which is 7k2u. Now if you're doing modeling and generally everything that you do with Haddock, it could be, especially for structures that have high conditions that you might have multiple conformations of some sidechains. And then the software will not know what to do with that because you have a sidechain existing in two conformations. You have to do or inspect the file and on which conformation you want to use, or you can automatically basically extract only the most conformations and we have developed the PV tools for that. It's also available as a web portal, so you can run this as a web directly on a server. But if you install it locally and we use it all the time, quickly run the command. So this is going to select only the most populated conformations, verbal occupation. We keep on the pity files. We want to remove the crystallization junk. So these are those ones. Water and then we generate a template. Actually, copy paste this comment just to show you that this is not junk, but it's my directory except for the data. And if we look at what's the difference, so this is not a template. You see that I don't have any crystallization junk. I still kept the cofactors and I have my template ligands here, which is not the ligand we want. It's the most similar ligand that we selected. OK. Now we need not to generate this shape. So we need to transform this information to a shape. So we're going to... So the name of this VU7. So we take that one and here we have a simple script given to shape, which basically rename the atoms of the ligands to what we call bead fake atoms. So what is in this shape file? So for other processors, you see the residue name is SHH for shape and the atom name is shape. Each atom is a separate residue number, shape which consists of 30 atoms. If I look at the final basically what you are seeing now, it's basically a three-dimensional shape. It's still in the orientation that we find in the template. Love to maintain these orientations in the document. That's a love of... And of course, we want to use the template, the receptor for the dock. We want to have the ligand in there, otherwise the docking is not going... So I want to remove the ligands and now what I have here. So I have the template that... Which is where I have my shape, which is ready for docking. But what I don't have is my ligands. So now we need to generate formations for the ligands and we need to worry about conformational changes. Okay. So we go to it for this, starting from a smile. And if you have installed everything properly as instructed in the setup of the tutorial, you can use this script, a Python script to generate the conformers. So I'm going to... The script is provided with the data you download. The smile string of the type where docking is provided in data. So this is the command. So I'm going to do this conformer generate life. Always key. I just run our script and what has been generated is this file conform. So what is in this file? So we have now an ensemble of conformers. So you see model one, the name of the terms, model two, model three and model. So we have a number of conformers. In this case, we have 16 conformers that were generated. Formers that generate might depend also on the version of party because it's changing. So if I look at this, what's in there? No, that's conformer. It's interesting. This unused stick issues here. Try to load the one from the data directory. It's a relation issue, but it seems that Paimol doesn't display it. So here you see the one that we pre-generated. And so you can realize that there's not so much conformational viability in it, but you can see maybe around those bones here, there may be some variation in here. And if I play Paimol as a movie here, so these are the 16 conformations in this one. And basically you see the amount of viability. So each conformation is like so instead of docking with one thing, we give to add up this ensemble of conformation. I'm just trying to figure out wrong with the one that we just generated. Or it's the, well, the connect statement seems to be messed up, that are the key generated. So if we remove those, then use them all this, displaying those properly. This is the one that I just generated. So we have, actually, this is the lab run, okay? That I forgot to write in the tutorials. Everything is fine. So we have now an ensemble of conformations. We have the shape and we have the template. So now we need to restrain, to generate the restraints to guide the docking process. So basically what we want to do is to, so there are fewer atoms in the target compounds than in a, so we're going to define the restraints from the target to the shape. If your shape is smaller than the ligand that you dock, so you have less in a shape than the ligand that you dock, then you should rather define the restraints from the shape to the ligand. If there are questions at any time, just do pull during the chat. I see we still have a minute, okay. So we can generate the restraints for the shape. So this is the format that Hadock takes to define restraints. It's a CNS slash expo format. So from all atoms of the ligand, and we'll have chain IDB. And the shape is basically defined a distance restraints to segment S. And this is an ambiguous restraints. So it means that the home should overlap with one of the atom of the shape. It can be anyone. We don't specify which atom because we... So this is a set of restraints which we're going to dock. Since the template contains two cofactor, it also generates restraints to maintain those cofactor base because they might start moving as well. So this is recommended in Hadock that we don't. So this would be this one and basically two file, one file per type of restraints. And now I'm combining those two files into one. And what you need, these are now very specific distance restraints to atoms. So this is a classical distance restraints and you'll see that there is a... So this is the difference that as we measure in a PDB file and it has an upper of one extra in this case. So we just want to basically keep the ligand in position. So now we have everything we need to set up the docking. So we have the template structure clean. We have an ensemble of conformations for the... That we want to dock and we have restraints to get the docking. Let me close some reaction. So now we're going to go to the dock server provided you have registered. Or we go to Hadock 2.4. So now you are in the Hadock portal. I'm already logged in the portal. So I can log in otherwise you will have to log in before you can do any submission. There are different interface to the server. The regular submission interface is the submit or if you go here, submit a new job. Refine an interface. If you just want to refine a complex or you don't docking, you just want to refine a complex. And as you will see when you get the file, you get there's a JSON file created that comps the parameter on your data. And if you want to repeat the same docking run, you can resubmit that file or you can add or change some parameter and submit it again. It's useful when you want to do multiple runs just changing a few parameters, especially if you have to change a lot of parts, you don't have to go through all the menus again. From here, you find, okay, there's a registration button, some missions to our tutorial page, the bottom where you can ask or search for answers. Someone was asking about the filters. You can find another cofactors, the modified amino acid, you can find the list here. Now, someone is asking a number of molecules. So for virtual screening, it's not what you want to do because you will have one. If you have praises, you can automate, if you go to the git repository, look up the paper, you can automate a lot of the search, of course, but I will not recommend these protocols for virtual screening. I think they are better, small molecules through two. If you want to screen one million of compounds in principle, the docking will do here if you have say a full node would probably complete in five to 10 minutes, depending while they are pre-processing, but it's not going to be taking a huge time, but 10 times one million becomes a large number unless you have one million nodes, give resources, and then you can do a docking 10. Okay, so that's more, I think you should see this protocol more as a way of, if you are really further down the pipeline, you want to generate maybe more reliable models to start doing structure-based drug design, things like that, that's where it will be useful. So for virtual screening, use the classical small molecule tools. So we're going to submit a new job, to give a, so it's a pre-school shape, whatever. You should not use a character to Linux, like space and dollar, however, we'll tell you that that's not a good idea if you do that. Molecules, I mentioned how dock can go up to 20. In this case, we have three molecules. We have the protein, the receptor, we have the ligands, and we have the shape. So I have to set it to three. You do that basically, there's a third molecule input field that appears. So we have molecule one, molecule two. And then for each molecule, you see the server gives you an option to download it directly from the PDB. That's probably an option that we should from the server, because if you download it from the PDB, all the cofactors, all the crystallization, all the water molecule will be kept. And this might give you trouble in the dock. So it's always recommended that you look at this before you do the docking. So we are submitting it. We are going to change, it's clean, but you could select a specific chain. So let's find this file. It's in, it was my template that I generated, upload. And this is a protein, or it's a protein containing small ligands, actually in this case cofactors. What I want to do here is to fix the molecule in its orientation, because this has to be done also for the shape. So the shape that we derive from the sector, and if we start moving around or doing random rotations prior to docking, then the shape does not match the binding site anymore. So in this example, we fixed the molecule. So this is basically number of molecule, fixed molecule. Next step is to input the ligands. So for the ligand, we're going to use this ensemble of conformers, and we have to tell the server that this is, so this is molecule two. It's for this one, molecule two, select. So these are my conformers. This is the one, the clean one. This is a ligand only, and you see here all the different types of molecule that we are supporting. Unpeptide glycans, I didn't speak about oligosaccharide, we do support oligosaccharides, nucleic acid, protein, nucleic acid, and shape. So this one is a ligand. One, we don't want to fix it in space, because we would love to dock it. And the last one that we're getting, so we're selecting now the shape, which is here, upload, and this is a shape. And this is a shape. And it's fixed automatically and importantly, the shape segment ID or chain ID for docking, which is called S. It's important that the restraints are defined, if you see again your shape, sorry, well, it's important that the restraints that you define match the chain definitions that you give to the portal. The shape as chain S, the ligand as chain B. And if you don't match these, these restraints will not be read properly. So this is, you can also see that you have an option here to define the charge state of the termini of your protein. Often what you have in the PDB is not the full sequence that was used for the experiments, but only the part that you, the density, meaning that the end in a PDB file match the real end of your molecule. And for that reason, by default, we have uncharged termini, but you can override that. I'm ready, we come next. So now some validation took place, HADOC calculated the protonation state of, if you find CCD and it has generated parameters for the ligand, you could select here residues that define the binding pockets, but we don't need that. What you can see here, what is also is a CCD and protonation state. So HADOC has identified that you have all those CCDs system and actually they are defined as uncharged. So this is, they are free states for CCDs that can be possible. One is the, so they are protons on the two nitrogen on the ring, but they are neutrally CCDs. And in a neutrally CCD, you can put the proton on the delta position or on the M position. And in this case, based on the mold property, output or the HSTD enough, is E state. Now we don't do anything here. So we go to the next stage. Since things were ligands, we have automatically change parameters on the portal to ligands recommended parameters. So we have, so we're now at seven, okay? So like input parameters is what we just did on step eight. So now we need to upload the distance restraints to the set. So there are different ways of uploading restraints. So you have ambiguous and an ambiguous, but these are just names. Since we have two set of restraints, we're going to use as ambiguous. I'm going to give the shape restraints. You could even combine those files into one, C, E, B and S, upload. And as an ambiguous restraints, I'm going to give the cofactor restraints to maintain the cofactors in their proper location, upload. We don't need restraints or small molecule docking. We recommend to keep all hydrogen atoms for protein docking by default, we remove those because had a united atom model. So we don't need the, it's only removing by the way the non, so we keep all hydrogen that we have a partial charge. I told you that to deal with bad data or false positive we'll exclude 50% of the info. In this case, we don't need to do that. We have the shape, which is quite good. So we turn off this option. Note that again, that depending on your access level, when you're the first time, you will get easy access level and you might not be able to change all those parameters. To change more advanced parameters, you should request for higher level access and you can do that in your own registration page. So we have defined the list. So this is all good. So now we want to change the sampling. Okay, so we have 16 conformations of our ligand in this case. So we're going to sample each conformance. So we want to generate 20 times 16. So this is 320 models for working. So I go to the sampling. So the server allows you to modify probably 500 different parameters. If you don't know what you are doing, don't start changing parameters. Sometimes when you hover about parameters, it will give you some information on what it is. So numbers to doc, default thousand, which will be 10,000 sampling. So we here do 320. And since we're... So something that the server does by default also, when we generate a model after rigid body docking, we automatically rotate by 180D solution and re-minimize it. Because we realize that often you get symmetrical solutions that are difficult to distinguish. In this case, since we have very specific, you don't want to do that. Also for small molecule docking, we recommend to stop the protocol after flexible refinement stage. So you don't want to do the final refinement on that off. That's this option here. Also, if you remember the view of the binding site, it's right binding site. So we need to penetrate inside the protein to facilitate that we scale molecular interactions for the rigid body docking to only a per million of the total. So this is in the interaction energy and interaction parameters. Scaling of interactions for the rigid body docking, zero one. And because of that option, we also need to change the scoring function because this will generate generations of more clashes. For this reason, we should put the Van der Waals energy term at IT zero for the scoring to zero so that we don't penalize clashes at this stage. We refine the way anyway. And in this case, so we know to do any analysis of the results but basically setting the analysis to none what we're going through, what the server is going to analyze is the top 10 model by the docking. So we don't do any clustering of solution. And with that, we are ready to submit. So I can click on submit button, not COVID related. Your job has been successfully processed. And now, okay, so we have about 10 minutes left. So as the job is going to start running, we're going to press based on the number of models. I mentioned this one file, which is always recommended. I'm always recommending it you to save it. It's this Chasen file. So we can have a look at it. So this is a text file that contains all the parameters that takes. So there are many more parameters here than what you have seen through the interface. But for example, we should be able to use the number of structure which was 320. Here is another parameter that we changed in a portal. This was this intermolecular interactions for IT0 that we set to one. If you say, well, this was not a good choice and I should have set this to 0.01, you can delete this file and upload it to the interface of the server file submit interface. You don't have to re-enter all the data in all the... And what you find in there are not only the parameters, you find the STDN, by the way, but you also find the original PDB file set you submitted. So this contains all the data that you... So this is the protein. Then you find here the models. So this is the sum of models for the ligands. And if I go down, I should find the shape. The shape is here. So you can also extract this information out of the file. And you also find the restraints that we... The ambiguous restraints to the shape and these are the restraints that were defined for the co-factor. So it's really a same file. It's a good reference of what you have been doing. If you were to publish something out of the work that you have been doing, this is a file that you could provide as a supplementary material. Okay. So we are here running... So let's move to the final thing, which will be the analysis of the... So we're not going to wait for this one to finish. Actually, some before that did complete. It's here. This one was submitted a little bit before and you see we have the result page. But this is also available from the tutorial. So if you click on the pre-calculated page here, I'm going to open it in the new. So this is accessible to all of you, also if you're not registered. So some general information, citation of HATARC, citation of the projects that are finding all of this, including the BioXL, if you are happy and unhappy with our portal. So we always like feedback from users. And then you... Since we set the analysis to normal, this is by default, it does the top 10 model. So what we have here reported is the top 10 model. So cluster one will be the first, the top-ranked model. Cluster two is the second-best-ranked model and you see the scores. So the scores of the second and first one are, in this case, are the first digit. And then while someone asked about how is it compared to the methods, I just showed that at the beginning of the tutorial. And it's also in the paper, actually. So this is the story here. What we are comparing here is this protocol that's bond docking with the other methods. We are doing fully unbond docking, we're generating conformation from strings. If you look at the best model that we generate in terms of quality is the shape model. And this is what surflex gets, but it's using the formations from the crystal structure of the complex. So that's a fair comparison, but still you see that in terms of quality, we are doing better than many others, starting from unbond structures and from, while here the other starts from crystal structures. And if you look at the top one, we are 30%, we will shape protocol, bold and glide do slightly better. But again, this is bond docking. There is no conformational changes while we are doing bond docking. Again, it's not a fair comparison, but it's very competitive with the other method considering that we are doing unbond docking. We don't have the data for the other methods. Back to the result page. So here you have basically 10 models. You can even on the portal directly visualize the model. Let's move it back. So you see the shape is drawn here, but the ligand is not shown, so the line representation to start seeing the ligand is a quick check of what's happening there. And if you scroll to the bottom of the page, you have a vision of the results. You find four versus a fraction of common contacts for small ligand, this is not so relevant. RMSD at the interface, this is more relevant. Where you see the top 10 model and you see all models, you see there are some models that have a very high clashes. You cannot really see what's happening here, but the active plots, so you can zoom in if I want to know, okay, what's happening here. You can zoom in, the Hadox scores and the RMSD interface between the ligands. You get the same for this elevation energy, von der Waal's energy, Electrostatic, you can see here that there are different sets of solutions from others. And you also have here a representation of the results. Again, because of those that have high value, these plots don't, fantastic, but you can zoom in on what's happening here and you will get a better view of things. Okay, so what we can do is to visualize the structure and actually we have been doing compared to the reference complex, okay. So the server allows you to download all the clusters. So you can download the here. If you download this, you get models from the rigid body docking and the refinements and you will find here 320 models per stage. If you're just interested in downloading the cluster and click on this, you will be able to download all the cluster files, okay. And now we can basically compare those to our reference which we know the answer. So I'm going to use PyMol to do that comparison. So I'm going back to my work directory. I need to extract, okay, I've extracted. So I just generated our 10 clusters and now I'm going to basically load those 10 cluster PyMol with the answer. Just to see what world we do it. So I'm going to remove anything that we don't want. I'm going to align all the models to the protein to the receptor. In PyMol you can copy paste part of the PyMol command window. If you're not familiar with it. The arm is the only on the protein side. I'm going to hide the shape, zoom on the ligand. Okay, so now we see here in green, the top 10 that PyHadox using this protocol and in white reference structure, basically the complex that we are trying to reproduce. And you see that, so the, you can see that all top 10 models are very, very similar. We have a very nice fit here and there's a little bit of a difference here. And we can do much in terms of arm SDDCs, but basically the work extremely well. I hope to have convinced you of that. And this is another view that you find here. Basically, now you can calculate the arm SD of the ligands but basically, and we use again a hardy kit to do that. So we fit on the protein interface of the binding side interface and we calculate the ligand arm SD compare. So you can only do that if you know the answer. So do that, if you don't have profit, which is a useful software to do arm SD calculations, you can do the fitting and then export the ligands. So I'm going to use because I have everything installed on my laptop and you see here. So basically what hardy kit is telling me after fitting on the interface in is that the arm SD of the ligand, the reference structure is 0.74 angstrom, okay? So the difference, this shift that you see here in some orientation of ligand is 0.74 angstrom. So I think that's an extremely nice result, okay? So as an illustration of basically this shape-based protocol, so what we've been doing is going quite fast over part. If you are interested in doing more protocol, everything is described here as well. This was very much ligand specific, showing you the latest and most performant. We have our small molecule now. If you are interested in protein modeling, you have a protein. So if you're interested in peptide modeling, look at the protocol with the tutorial, which is part of this master course. So you get a lot of results there. See, okay, one question, what would score? So score are comparable between different runs, but they will depend on the data that you put in. The runs of the same system, you should be able to compare the scores. You cannot move the restraint energy from your comparison if you want. But the scores, what the scores of different ligand, the same protein are probably okay, but these not correlating usually very well with binding affinity, okay? So you have to be careful there. What would be a good score? No value which tells you what is a good score. I know that Autodoc, for example, generates bio energies in K-Calpemol, Delta G, that will score of arbitrary units. And you cannot really, if you're comparing the score, another protein with another ligand, you cannot, one is better than the other one, binds better. Of course, this is energetics that we're calculating, but if you claim this is correlating with binding affinity, this would be a false claim, it's not. And any software that tells you that is putting a claim, which is probably not validated. And if you want to know more about that, go look the D3R Grand Challenge, there were also a challenge of predicting binding affinities and ranking ligands, and you would see that it's very hard binding affinities, or you have to do fancy free energetic methods, assuming that your binding poles are correct in first instance, so that's much more work. So be careful in interpreting docking score for any software in terms of binding affinities. Are there any other questions? It's exactly 430. I realized it was a formation in the short time, but I guess you got a feeling and do. Yes, and we are having a recording of all the sessions put on the web later, so people can watch them at speed again. Okay, so if there's nothing more, thanks to Alexander and Rodrigo in the chat also for the very nice session. And everyone, as you know from the, now that we've covered HPC in general, your dynamics, range calculations, and like your docking, tomorrow we'll continue with looking quantum mechanical aspects of your simulations. So it will be a very nice follow up of the, thank you everyone, have a nice evening and we'll talk tomorrow. Okay, bye everyone and enjoy the remaining of the school. Yes.