 And good afternoon. Welcome to the third webinar in the BioXL webinar series. My name is Adam Carter, and I'm very pleased to have with me today Adam Aspetal, who's going to be talking about atomistic molecular dynamics set up with MD Web. Before we get started with that, I just want to give you a very quick overview of the BioXL project. We'll have more than five minutes just to give you a little introduction as to what we're doing. And then Adam will speak for 40, 45 minutes or so, and we'll have some time at the end to take questions. So we can take all the questions at the end of the session. Before I go any further, I should let you know that this webinar is being recorded. You'll be able to see the recording on YouTube, and this will be linked from the BioXL website a few days after this webinar. So you'll be able to catch up with all the details there. And you will also be able to see details of past BioXL webinars as well. So the BioXL Center of Excellence, it's a new center of excellence funded by the EU. And we're a center of excellence for computational biomolecular research. The center is built on three main pillars. Essentially, we talk about three pillars which describe the BioXL Center of Excellence. The first one is excellence in biomolecular software. So on this slide we mentioned Gromax, Haddock, and CPMD. These are some key codes in the biomolecular research community. What we want to do in this project is to improve the performance, efficiency, and scalability of these codes. We're lucky in that the lead developers for these codes are partners in the project. So we have a direct line to these simulation codes. And it means that we can also support and feedback into the codes development. Another aspect of BioXL is excellence in usability. So as well as having these three codes that we're trying to improve, we want to think about the codes in the context of wider workflows. So what people are actually doing with them and trying to make those workflows work more seamlessly, more straightforwardly. And as part of that in the project, we are trying to compile workflows using some standard workflow platforms like the ones listed at the bottom here, which will bring together the codes along with the codes that I've just mentioned on the last slide, along with lots of other tools and services that are used in the biomolecular research community. And as well as other tools, we're also considering bringing in data sources as well. So trying to build an integrated platform to allow people to make the most of all the different tools that the center is supporting. And finally, we are interested in, as well as offering the supporting the software side of things, we're trying to build competence both amongst academics and industry users. So we're trying to promote best practice and train end users to make the most of these pieces of software and to make the most of the systems and the environments that we are making available. So we're targeting both academic and non-profit users here and industrial users as well. And the project is also reaching out to independent software vendors, academic code providers, and academic and commercial resource providers as well, so providers of services and hardware. So that's a kind of higher level description of the kind of things that BioExcel as a center of excellence is trying to do. One of the key things that we have for our users are interest groups, and we're launching initially these six interest groups that you can read on this slide here. One interest group that will possibly be of particular interest to some of the people on this webinar is the Biomolecular Simulations Entry Level Users Interest Group. So we'll be sending an email after this webinar to invite you to join these interest groups if that's something that interests you. Alternatively, you can find out more about them on our website. The interest groups will have mailing lists, access to the forums at ask.BioExcel, and we'll also be able to provide code repositories, chat channels, video channels, any other sort of collaborative platforms that are of use to the people in the interest groups. So hopefully that gives you a very brief overview of the center of excellence, and what we're trying to do. So now I would just like to hand over to today's presenter, and I'll just give a few words to introduce Adam. He's a postdoctoral fellow in molecular modeling and bioinformatics unit. The MMB hosted at the Institute for Research and Biomedicine in Barcelona, and he's a computer technician for the Spanish National Institute of Bioinformatics. He came originally from a computer science background, and he jumped into the bioinformatics world and got trapped there by the fascinating field of structural bioinformatics. After two years, he joined INB, and he's been working there for more than 10 years. INB, amongst many other projects, has recently joined Elixir, a large European project which is building a sustainable European infrastructure for biological information. And he's also been involved in several different projects at IRB and the Barcelona Supercomputer Center. While working as a bioinformatician, he got his PhD in biotechnology from the University of Barcelona with his thesis, High-Throughput Computational Studies of Macromolecular Structure Flexibility. And since then, at INB, he's developed a set of public web servers and databases related to macromolecular structure flexibility, including MDWeb, which is the main topic of today's presentation. So I'm very happy now then to hand over to Adam. I will make you the presenter now, and you can take over for the rest of the talk. Okay, hi everybody. Can you see the slides? All good, Adam. Carry on. Okay, so thanks Adam for your presentation. As Adam said, my name is Adam Hospital. I'm working in IRB Barcelona Institute for Research in Biomedicine. And I'm going to talk about atomistic molecular dynamics setup using our web tool called MDWeb. I have divided this webinar in three main blocks. In the first one, I will give you a brief introduction about the importance of molecular flexibility, about a theoretical method called molecular dynamics, and its limitations. I will briefly introduce three different platforms developed in IRB, Model, MDMovie, and MDWeb. And then I will move to the second part, where I will focus on MDWeb, that it's a platform to run molecular dynamics on web. And in the third and last part of the webinar, I will just introduce other web services, services that we developed at IRB, all related to macromolecular flexibility. So starting with the introduction, I guess that if you have registered to this webinar, you are all aware about the importance of molecular flexibility. But just to reinforce that, I would like to present just three nice examples about this. The first one, if you can follow the pointer, is a cross-talking experiment in one particular protein that it's called acetylcholinesterase, where we extracted the ligands. Here you have different PDB codes that are different conformation of the same protein. This is the same protein, but different conformation. So we extracted the ligand from this conformation, and we tried to cross-talk these ligands to the same conformation again. As you can see here, the docking results were really bad. Blueish squares are bad results, reddish squares are good results. So this is telling us that the differences between the different conformations are very important for the docking of a ligand. And even if we look at the diagonal here, that is when we try to dock the ligand, we can extract it from one particular conformation to exactly the same conformation. In some of the cases, we also got bad results. So this is telling us that there are some structural rearrangements after docking of the ligand. So again, flexibility matters. In the second example here, it represents how some properties like protein channels in here, in the left part, or drug cavities in the right part can only be seen when looking at the dynamics of the protein. So certain movements of the structure can make internal channels accessible from the exterior like these ones here, and new cavities like these ones here can be formed in the surface, so allowing the recognition of small molecules by the protein. And finally, in the third example here, this is a typical example of a gate opening molecular switch. So here residues in the protein can act as a molecular switch, a door that opens and closes, letting small molecules diffuse inside the proteins. So in this particular example, this is a study of a truncated hemoglobin. You can see here the events of opening and closing of this door in the wild type protein, and these events almost disappear completely upon mutation of the key residues involved in this gate. So molecular flexibility is important. So that we should not look only at the static pictures as we have them in the PDV, in the protein data bank, but we should look at the dynamics of the protein. So how these proteins move and the flexibility information. Unfortunately, with experimental techniques, it's difficult to obtain this experimental information. The flexibility information is difficult to obtain by experimental techniques, although we have some of these techniques such as NMR studies that can give us important insights about these dynamics. However, when we are dealing with more complex systems, such as big proteins or even macromolecular complexes, these experimental techniques are still not able to extract this flexibility data. So that has increased the importance of theoretical techniques, such as, for example, molecular dynamic simulations. Molecular dynamic simulations is the most widely used theoretical technique to obtain this flexibility properties. And it works solving the classical equations of motions where forces, acting on every atom, are obtained by deriving these equations that are called force fields, where potential energy here is reduced from molecular structure, bonded terms, so bond angle and dihedral, and also non-bonded terms such as van der Waals and electrostatic energy. The simulated system could be represented at different levels of detail, being the most used the atomistic representation, this one here, where we take into account all the different atoms of the structure, all the coarse grain models, like this one here, are becoming more and more popular. In these cases, we gather together in single bits more than one atom. With the atomistic resolution, we can reach time scales from picoseconds to microseconds, whereas from using the coarse grain resolution, we can reach microseconds to milliseconds timescales. When talking about molecular dynamics, we must mention the Nobel Prize in Chemistry in 2013 that was awarded to Martin Karplus, Michael Levit and Ariel Warshall for their development of multi-scale models for complex chemical systems. They are considered the fathers of molecular dynamics simulations, and I recommend you to take a look at the Bioexcel EU webpage because we have an interview that will be published soon in the webpage. We have an interview to Michael Levit, so take a look at the keep an eye on the Bioexcel webpage. Why don't everybody use this molecular dynamics technique? Mainly because it's usage limitations. The main ones are these three limitations. The first one is the large computational resources that it needs. Typically, we use hundreds or thousands of processors to run a single simulation. We have also uncertainties with the four fields or the parameters that we have to run these theoretical techniques. We still don't know if they are accurate enough to reproduce the reality, and still today new four fields are being developed to solve these problems, the problems of previous four fields. The last one, maybe the most important for us, is the high level of expertise, the steep learning curve that it has to start running, to start doing molecular dynamics. At the end, if we want to go from a static structure to a dynamic one using this theoretical technique, the first thing that we encounter, it's like a wall in the middle of the road. We need to climb the road, we need to jam this, but we need to overcome these limitations. Let's be a little bit optimistic and let's go through these limitations and see how we can overcome them. For the first one, for the computational resources, these macromolecular dynamics theoretical techniques, they are taking profit of the so-called high-performance computing HPC system. The recent MD codes like AMBER, GROMACS, NAMDI, they typically allow running simulations using hundreds of processors. For example, from these powerful supercomputers like this Marin-Oson that we have at the Barcelona Supercomputing Center. Another way to overcome this limitation is, of course, if we don't have access to HPC centers, we can use distributed computing initiatives like these ones, where simulations are run using personal computers from volunteers around the world. Some examples of this is Folding at Home, started at Stanford University already 15 years ago, or a more recent example, the GPU grid here in Barcelona. They work using the idle time of the computers, of personal computers around the world. For example, when you have the screensaver in your computer and your computer is not processing anything, it can run a piece of a simulation that then is gathered together and joined together, building a single result. Another way to do that, and maybe the perfect way to do it, is build a machine completely specific to run these molecular dynamics. That's what people in DE show research in New York did with this Anton supercomputer. Of course, this Anton supercomputer is, unfortunately, a private resource, and only a few of these machines are available for public research. But it has pushed away the boundaries of what can be simulated using MD. So it can produce simulations times in the millisecond timescale, so timescales that had been never explored before. But back to the reality, what has really changed completely the field of molecular dynamics is the appearance of the GPU card. So did this hardware architecture that were designed specifically to accelerate the generation of frames per second in the field of 3D computer games? But they tend to be optimal, they fit perfectly for computing MD simulations, just because they have the ability to compute thousands of operations in parallel. So most of the MD codes existing today have been adapted in the recent years, some of them with the complete pre-design of the code, to reach performance similar or even better than hundreds of common processors working in parallel just using one of these GPU cards. And of course you can buy one of the GPU cards and have it in your own machine. So all of these HPC systems, together with the improvements on the MD codes, and especially the possibility to run these codes in these GPU cards, these have helped in not just overcome the computational resources limitations that I presented before, but also to popularize the field of molecular dynamics simulations. So for the limitations regarding the uncertainties of the force field parameters, we should also be optimistic here because we are getting closer to the results obtained with high resolution techniques such as quantum mechanics. So to illustrate that, I will show you just a recent example of a reparameterization of a nucleic acid force field. This is a force field specifically designed for nucleic acid simulations that we developed here in the MMB group in collaboration with the Barcelona Supercomputing Center. So we already published one of these force fields in 2007, almost 10 years ago. And so we found a problem with this problem a couple of years ago when we were able to reach the micro-second time scale with improvements of the new hardware as I presented in the previous slide. So we started to work on a correction of this force field, but in this case we focus on the parameters that we know that were wrong, so sugar-packeting, for example, epsilon-theta and guide torsions. And we managed to publish that after an extensive test so many, many different simulations, many, many different systems to be sure that this force field is working well. And here, the most important thing about this slide is here. This plot here where you can see this is the old pattern BLC0, this is the new one, and this is the quantum mechanic result. So you can see that we are getting closer to the high-resolution data. So we should be optimistic about that. And of course, we are still working on these force fields and we are improving every time we encounter a new issue. And this is a particular example for nucleic acid, but of course there are different groups working also on protein force fields. The last limitation is the level of expertise needed, the steep learning curve, and this is one of the most important for us. And to illustrate that, I have plot here just one example that is the typical pipeline that you will find in the books that explain molecular dynamics, and this is just a setup, this is just a preparation of a protein structure as you have it in the PDB to go from this static structure to a completely prepared system surrounded by waters with counter ions to neutralize the charge, etc. So a completely system prepared to be used as an input for molecular dynamics simulation. So you can see that it has a lot of different steps. So basic steps, more complicated steps like equilibrating the system, we are using different programs, in this case, dilip and sander from amber package, but also CMIP to neutralize the system. We are using also scripting mixed here to know if the structure contains a ligand, we are using a database of parameters for the different ligands, etc. So it's really complicated, it's not easy to start using the ND if you don't know how to run all of these different steps. So how can we overcome this level of expertise problem issues? Before answering that question, let me jump a little bit on time back to 2010, actually 2006, that was the time when we started a project in the group that is called MOLE, Molecular Dynamics Extended Library, where we wanted to simulate a representative set of structures taken from the PDB, so a representative set meaning that they shared less than 90% of sequence similarity between them. So we ended up with 1600 of the structures and we simulate that with different programs, with different force fields and to manage to publish that and we published that four years after that, after the beginning of the project in 2010, we built a web interface connected to relational database to make all the results publicly available. So I'm not going into details about the model project, but I just want you to realize that what we couldn't start 1600 or 1800 simulations manually, we need to build an automatic MD setup run and analysis pipeline to do all of these simulations. So that was the starting point for our project, that was called MDMovie. It's called MDMovie because it's using a library that it's called Biomovie and it's a library to build web services. So what we did is extract all of these small pieces here, so we split the pipelines into different pieces that it has and convert these pieces to web services, independent web services, so little pieces of code, of programming code, that do small things like these ones here. So using different programs and with that we could use different MD packages and we could join together the different web services building any kind of workflow that we wanted. For example, sorry, I forgot to tell you one thing and it's that this Biomovie library allowed us to not just build these web services, but also to add semantical information to these web services. The inputs and outputs of these web services were not just XML files, they have biological information. So for example, we have an object that is a PDV and we know that the PDV object contains the information about the atom, the residues, the X, Y and Z, Cartesian coordinates, etc. And we have another, for example, another object of the ontology that it's called Gromax, MD Trajectory and we know that this object contains the XTC Trajectory, for example, the TPR binary file to run the simulation, the GROW file, the top topology, etc. So we have objects with information that allow us also to find out which web services can be run from one particular object of this ontology. We'll see that later on. But we could build a set of workflows from these web services. This is just an example, generated topology for Gromax with this amber force field. Here's just an example to illustrate you that we can also mix scripting, these boxes here are scripts, these one are web services and these are the ontology objects that I was explaining in the previous slide. So we built a set of workflows with different programs, amber, MD, Gromax and for different pieces of the workflow. So we can just generate the topology if we want. It would be this one here. We can run a setup that is generating the topology, that it's fixing the side chains, adding the hydrogens and in the setup process minimizing the hydrogens and minimizing the structure. The setup with solvation will be the same, but adding the box of solver and also structural waters. And the complete setup will be all of these and an equilibration to equilibrate all the box of and the protein. So obtaining a final system prepared to run the micro dynamic simulation. And the last part after having this MD movie set of web services and workflows was to have something to ease even more the usage of this of this infrastructure. And for that we built the web interface that is called MD-Web where we could directly access to all of these easily to all of these workflows and web services. Now we'll move to this MD-Web in the second part of the webinar. This is how the main page looks like. This is the link of the web server and just to tell you that you have here a login a step I recommend you to register it will take you just a minute or less than that and it's important because you can work as an anonymous user but your workspace will not be persistent so if the PHP session expires or you close the browser all of your work will be removed. So just register it just a minute and you'll be able to start using the platform. So there's two different ways to start a new project in MD-Web the first one is working from a base structure so doing a setup of the structure and the second one is working from a trajectory so analyzing the resulting trajectory. So we start with a structure one and the first step and that is something that is not implemented in the web services is the structure checking. So this is the first step and it's a really important step because the correctness of the input structure is crucial for a simulation. So if we take a look here we can choose between different so the checking page is divided into three main blocks the first block is something that you can choose or you can fix. So for example we can choose models that you are interested in simulating you can choose chains also atomic alternate locations and you can choose between different AIMI assignments and I will show you that in a live demonstration now in a minute the second part is information is warning that the checking page is giving to you so like missing atoms or missing residues meaning that your protein is not complete and that's important also for the dynamics and also the atomic clashes that your structure can have. And the third part is part about the ligands so the checking page is telling you if your protein has known ligands known meaning that we have the parameters for this ligand in the database from our model simulation database or unknown meaning that we don't have the parameters so you can upload if you have it or just generate the parameters so now let me try to show you this here in a live demonstration here is how the checking page looks like this is a transcription factor protein bind to a look like acid structure so the first thing here is that if you go to the PDB you can see that this is the acid metric unit but the acid metric unit has two copies of the biological unit you see this is the biological unit one this is the biological unit two so if you are if you want to just simulate the biological unit you can go to the chain and just select the change that you are interested in so for example this one just select that and now you have the biological unit and you will simulate or you will prepare just this part of your protein so all of these are green checks and then here you have a red check you can take a look at this is a polar acceptor clash so if you put the mouse on top of this clash you will see where is the problem the problem seems to be that we have two oxygens that are two near to each other and here you can take a look at the other another problem that is an oxygen that is two near a carbon let me show you another example there is a ubiquitin and in this case the toy protein ubiquitin you have a lot of different analysis for this protein this protein has an amide assignment that may be incorrect according to our checking page and you can see here the problem you have two nitrogen that are two close to each other and actually if we go to the polar donor clasher here we can see that actually this is a problem this is a clash between two nitrogen but in this case we can fix this just swapping the nitrogen for the oxygen in this amide so we try to fix this, pre compute everything and now the polar donor clash has disappeared and we can go ahead with the simulation we still have one possible problem here we have two oxygens that are two close but in this case we can also here see that we have a possible hydrogen bond here between this oxygen and this nitrogen and another one here between this nitrogen and this oxygen so it's not a major problem because it looks fine so we can go ahead click next and start with a setup within the web so after the checking page this is the workspace you will have a base structure here and an image of the base structure and then if you click on top of the base structure you have different options here, different ballads meaning performing a new setup operation performing a new simulation or optimization performing a new analysis so we can start with race model or J model be with the log file if something went wrong download the results or delete this item if we click on operations for example we have a list of operations to choose some of these operations are web services these MD movie web services that I presented before and some of the operations are workflows so pre configured workflows with all of the different steps and web services that are going to be executed one after the other for example if we choose one of these workflows, generate topology for gromats and we put the mouse on top of the question mark and the web will tell you what is actually doing this workflow in this case is removing the crystallographic waters from the structure is adding the sidechain, the missing atoms the sidechains and is adding the hydrogens with pdb2jmx tool from gromax package in this case we can also choose a different force field and we choose this one for complex workflows like this one here that may take a while even hours we have this workflow progress report where you can take a look at the plot you can follow your workflow and you can see which of the web services are already done which of the web services are still waiting and which is actually running at the moment if you are not interested in running pre configured workflows and you want to build your own workflow you can also do that just joining different web services one after the other from a base structure we can clean the structure we can add the hydrogens generating a topology that is with gromax, this is just a particular example minimize the hydrogens, minimize then the structure solvate the systems or adding the box of water minimize the system equilibrate the final system and obtain the configuration files to run the molecular dynamics here you can see the different data types these are objects of the ontology that I was telling you before and the MD movie ontology PDV structure, PSF numDistractor, stop example topology structures etc so depending on these bullets here that are the objects and the web is able to identify which web services can be run using these objects as input so with this we can go from this static structure to this prepared structure to use as input to run the molecular dynamics the next step is running the molecular dynamics but of course this is an online platform we don't have machines to run nanoseconds and microseconds for molecular dynamics we restrict that the molecular dynamics run in this platform to 0.5 nanoseconds so 500 picoseconds the third point is that MD web is able to return the configuration files without running the simulation but it will build for you all the needed configuration files for you to run the simulation in your own HPC platform, GPU card etc so in this case for example I put here 10,000 picoseconds with a temperature of 300 Kelvin with an output frequency to write in the trajectory file every 500 steps so I will end up with 10,000 snapshot so one snapshot per picosecond and I can download all the needed files after running this web service the web service in this case again if we put the mouse on top of the question mark will tell you how is generated in the configuration file so in this case MPT ensemble and if you click here and more information it will open the help page that is an extended help page with the information about all the different workflows web services, ontology objects etc and you have also you have also here in this part here tutorials about the analysis setup and also run that you can follow step by step and setup or analysis of your trajectory so now we are in the analysis part starting the project from a trajectory of course again this is a non-light platform so we are stricter the upload to up to 100 megabytes of information but if you strip the water molecules and you extract a representative set of a snapshot from your trajectory you can build a trajectory with less than 100 megabytes try to upload it to mdweb that's what I did in this case so for example I have a trajectory of 23 megabytes this trajectory is a 10 nanoseconds trajectory which I extracted just 1000 snapshots stripping the waters and I was able to upload it to mdweb and run in this case for example an rmsd basic analysis and plot the analysis here this is from gromac so this is nanometers again if we click on base trajectory it will open a set list of operations that are web services for mdweb here you can see web services to convert between different formats compress the trajectory to pcz convert the trajectory to pdweb files to bin post formats here the format dc, the format etc we have getters, get the trajectory fragment get the trajectory snapshot, get habitat structure and we have basic analysis that are basic but very important are the first analysis that you do to see if your trajectory is behaving in the right way like the fact of the residue, radius of the iration along the trajectory, rmsd along the trajectory would be this one, or rmsd per residue this one particular service here is flexibility analysis that is quite general but this is general because it's not a web service it's actually a link to another server that we have implemented in IRB that is called FlexServe protein flexibility analysis in this case from coarse grain simulation but also from atomistic simulation so it is able to use trajectories from our model database also from our mdweb generated trajectories and it is computing a set of analysis particularly specific for proteins principal components, variance profile, b-factor landscape a lot of different flexibility analysis and it is able to show you these analysis in a graphical way also with interactive j-mol-uplets and it is also able to work from a static protein structure and it has implemented three different coarse grain dynamics to produce a trajectory, coarse grain trajectory in this case and run the set of analysis this flexibility server allows me to move to the final part of the webinar where I just wanted to tell you about four different servers that we have in IRB related to macromolecular flexibility two of them related specific to proteins and two of them specific to nucleic acids simulation so two of them are web servers to extract flexibility information, flexer and na-flex and two of them are molecular dynamics databases of molecular dynamics simulations model and vignasim so I have already introduced to you flexer na-flex is a web server for the study of nucleic acids flexibility so which is specific for nucleic acids and it is pretty similar to MD-Web, it is powered by MD-Web so you can run a set of molecular dynamics and run an analysis of a simulated trajectory and run a set of flexibility analysis specifically designed for nucleic acids and of course again you have representations, graphical representation of all of these flexibility analysis model is a database as I have introduced you before for protein molecular dynamics we have more than nowadays we have 1,800 different molecular dynamics simulations that were run at the moment in the Marinerston Supercomputer we have a database with all the metadata and analysis of these simulations and we also have all the static data all the trajectories in disks and you have a web server to see all the information of the analysis and the compressed trajectories and finally in the most recent one the most recent server is vignasim big data nucleic acid simulation database where we did exactly the same we store trajectories for nucleic acid simulation but in this case we moved to the non-SQL to the non-relational databases we used Cassandra for trajectory coordinates and Mongo for trajectory analysis and metadata and we also built a web a graphical web interface to have access to public access to all of this information and of course we use NAFLEX to visualize how we have already run to all of the simulations that we have in the database so with that I just put here the links of the different web servers that I have presented remember that this is reported so you can take a look at these slides afterwards this is our bioexcel center of excellence these are the molecular dynamics packages which all my group are modeling and bioinformatics and in particular to our group leader Modesto Roscoe and to the director of all the web servers here in IRB, use of visual P and now if you have questions I would be happy to answer Thank you very much Adam, that's great we do already have some questions in the question section of the room I just want to remind people that if you do have a question you can type them into the section that is labeled questions and we will try to answer them so the first question was from Founso Founso if you are able to use audio you can ask your question directly Hello Can you hear me? Yes, that's clear Thank you, thanks for the webinar I am a little bit of a rookie in this, trying to see if I can work with my work I have sort of worked with a predetermined structure I mean, biomology modeling to predict the structure of my protein Can I use this tool that you have talked about for checking protein structure, correctness of structure on the predicted protein Well, you can use this checking page that I was presenting in this webinar This will just look at the distances between different atoms, looking at the clashes looking at the missing residues and so on But in your case, you need to work with some program that looks for the energetic or something more advanced than just distances But yes, of course, you can just upload your structure and take a look at the checking page Okay, thank you And if one is to start these molecular dynamics packages, do they have what advantages they have over the other? I mean, for somebody who is going to start afresh You mean from the different MD packages Yeah, the different MD packages available Yeah, well, that depends So one, the main problem, let's say, your issue is that some of them need license, so for example the molecular dynamics program in the amber package that is standard or PMMD, you need a license to do that to use this In Gromax, for example, it is free, you can download it and you can use it, and in MD it is also free And then it depends on the HPC cluster that you have if you have GPUs, if you have supercomputer, for example or if you will run this in your own computer, so it depends Okay, thank you Yes, I think that's quite a difficult question to answer in a general sense, but I hope that's given you some starting points The next question from Marni Marni, if you're available to speak directly, I'll unmute your microphone. Hello, yes, we can hear you Okay, I'm just wondering how is coarse grain dynamics different from molecular dynamics? Yeah, it's just if I can go back, can you see the lights? Yes, I can Okay, so in the atomistic resolutions we are treating all the atoms here, you're seeing all the different atoms in a single bit, so all of these bits are treated in these formulas, whereas in the coarse grain simulation we gather together a set of different atoms like you can see here in a single bit, so we have less bits so the resolution is not so high, it's slower and so we can simulate larger molecules and longer time scales because we have less bits When is the appropriate use of coarse grain dynamics? I mean if you were going to recommend if I want to study something, when can I use coarse grain dynamics instead of molecular dynamics? That depends on the level of resolution that you want to achieve, if you are looking at the hydrogen bonds for example between one ligand or between the same molecule you need to go to the atomistic regime but if you are interested in global movements you can try it with coarse grain Okay, thank you very much Okay, thank you, we've actually got quite a lot of questions queuing up, I'm going to get through as many of these as I can we can maybe take the other ones to the forum later on if you don't mind just to speed things up I'm going to read the questions now to Adam, we can follow up by unmuting you if you have any follow up questions so the next question is from Haitham who asks can one add post-translational modifications to amino acids using MD Web and if so how? That depends if you have the library of this post-translational amino acid we have some of them in the model database so maybe if you try it with MD Web we'll give you this known or unknown ligand if it is known is that because we have the library if this is unknown is because we don't have it so then we can prepare the library parameterizing this but I guess you can find the library for this post-translation residue you can use it if you want it's a matter of uploading the library of this residue in MD Web Okay Haitham I hope that answers your question if you want more details please submit a follow up the next question is from Elvis Martis who asks is it possible to build lipid parameters files for Gromax using MD Web? It is not possible to build parameters using MD Web it's possible to upload the parameters that you built with another program or another tool and the web doesn't allow to produce parameters. Okay thank you Dira has a couple of questions I'm going to unmute his mic in a moment but before I do that there's a question from Debajyoti Ghosh who asks can I use MD Web to simulate protein bound to a small ligand? Yeah sure he says I should prod RG server to build topology and then use Gromax but it is tricky and laborious I mean with MD Web you can do that but you need the libraries you need the parameters for the ligand that's the main point and you can use other tools to do that but MD Web is not prepared to build parameters for the ligand but of course once you have the parameters you can upload the parameters and run a setup and then prepare the configuration files to run the molecular dynamics with MD Web Okay thank you Dira I have unmuted you you said you had a few questions could I ask you to limit it to one and maybe a follow up because we're a bit short of time Okay so my main question is everything in MD Web is based on BDB file can you build new molecules new systems? Can you be more specific? Yeah for instance let's say you want to build a system with a protein and nanotube something like that No MD Web is not prepared for that Okay do you have one follow up question or is that? Yeah My follow up question is about can you do basic setup coarse grain simulations with MD Web Yeah well in coarse grain simulation you mainly don't need a setup because in the ones that are implemented in MD Web that are three we have Bronian dynamics, we have discrete molecular dynamics and we have a normal mode analysis and we are using alpha carbons only as a resolution so we can extract the alpha carbons and directly start the coarse grain simulation so you can run a coarse grain simulation with just the alpha carbons of your protein in MD Web without running a setup before Okay Thank you Okay thank you so the next question is from Pankaj, I hope you don't mind I'm just going to read this out to save time, Pankaj Kumar asks Hi I want to see or compare the molecular states of a ligand with a protein and one of its mutant, may I know how should I go through with this computational molecular simulation approach, so how to compare the molecular states of a ligand with a protein and one of its mutants Are you able to answer that Adam? But I don't understand if the mutated protein, if he has mutated protein and he wants to study the different states of talking with the ligand, or is the ligand that it's changing, can we try to open the mic? Sorry, I myself was muted there Yes I've opened the mic Pankaj do you want to speak to Adam directly? Do you have a microphone? Or maybe you could type your question into the chat session if you don't have a microphone We can possibly follow this one up later on Sorry I don't really understand the question Sorry, my fault It's difficult to convey it in a small space Pankaj sorry about that, maybe we can pose that question on the forums later, or you can get in touch with us directly Arunab had a question So I'm going to unmute your microphone Arunab, if you've got a question you'd like to pose to Adam Hello Hello, Myself Arunam Naik Hi Myself Arunam Naik from India Ok please carry on Yes, I have a specific question Is it possible to calculate binding free energy of a protein ligand complex from bromax denoted trajectories in MDOF? Is it possible? No I'm sorry this is not possible We'll try to work on something like that but in the near future It's not still implemented in MDOF So another question is Is it possible that bromax denoted trajectories can I calculate the binding free energy in Ember? I guess you can but I didn't I didn't do it, I haven't done it before sorry, but I guess it's possible Ok, ok, so can I hope that in future in MDOF we can expect this type of calculation We are working in different directions but we'll add to the list Ok, thank you Arunam Thanks for your question The next question was from Mauricio who asked Can I prepare a protein with glycans added on? I don't know the answer If this can be treated like a ligand I guess it can be done but I'm not sure The thing here is that MDOF was implemented and was designed and implemented to ease the usage of molecular dynamics but basic molecular dynamics and here we are talking about advanced molecular dynamics and MDOF was not designed to that Of course we can try to update and try to evolve our software adding new functionalities but this was not the first idea Ok Adam, thank you very much We do have more questions in the room but I'm afraid we've reached the end of our webinar slot coming up to 5 o'clock now at CET So I would like to thank Adam again and Adam if you could move to the final slide I'll just say a last few words to sum up today's webinar Thank you all for coming along If you would like to find out more about BioExcel or as I say to join one of the interest groups where you can ask more of these kind of questions on the forum then please do have a look at our website and sign up there for the interest group for the information about the interest group so please do sign up if you think you might be interested Finally our next webinar is going to be on the 10th of June The subject is mutation free energy calculations with PMX from Bert de Groot who's another partner in the BioExcel project same time on the 10th of June So if you're interested in free energy calculations then it would be great to see you there Thank you all for coming along today There will be a follow up questionnaire that is sent out to help us understand whether this webinar has been useful to you We would be very grateful if you could take the time to fill it in It's only two or three questions Thanks very much and we'll hopefully see you again soon at the next BioExcel webinar