Bioexcel Webinar #95: DynaPIN: A Tool for Characterizing Dynamic Protein Interfaces

So, welcome to this Bio Excel webinar. Today's topic is Dynapin, a tool for characterizing dynamic protein interfaces. Uh just so you know this webinar is being recorded and we will also upload uh a copy of this later to the Bioexl YouTube channel. Uh during the presentation you can ask questions with the Q&A uh function. Uh the questions will be answered after the presentations. Uh if you still have any questions after the webinar, you can ask them in the Bio Excel forums. There should be a topic for this webinar there. And today's presenters, Etsy is a computational structural biologist specializing in structural modeling and biomolecular dynamics. Excuse me. She completed her PhD at Utrex University under the supervision of professor Alexander Bonman where she advanced hadock. She then conducted her post-doal research at EMBL H Highleberg with Professor Teresa Carlo Mango and Professor O or Soloya Barabas uh developing M3, a protein complex modeling tool capable of handling 20 molecules using sparse experimental data. Uh in 2017, uh Dr. Karaka established the computational structural biology lab at Ismir biio medicine and genome center and joined dou uh ail university. Sorry if I destroy any names here. Her lab develops classical and AIdriven approaches to model biomolecular interactions and their dynamics. She received an EMBO installation grant and uh was also the first assessor from Turkey in CASP 14 and CASP 15 for the assembly prediction rounds. Our lab's first online tool, Proton, a structurebased web server for designing inter facial mutations, gained significant attention from the scientific community with 50,000 users per year worldwide. Uh Bertin is a computational structural biologist specialized in molecular dynamics simulations and biomolecular interface analysis. She completed her PhD at the Ismir biio medicine and genome center and douut alil university I'm sorry I'm so sorry under the supervision of Dr. Eski Karaka as a post-doal researcher at IBG uh Bin is currently working on the development of Dynapin a Python-based package for time resolved characterization of protein protein interfaces. She's also one of the contributors of Dynabbench, a large scale molecular dynamics data set designed to capture interface dynamics beyond statistic reference structures. That was a mouthful. I'm sorry. But uh over to you now. So I'll stop sharing and hand over to you. >> Great. Thank you, Otto. Uh and no worries. It was uh all perfect. Uh so our institutions names are a bit difficult to pronounce. Um uh anyway so um I would like to thank first the Bioexile team for giving us the opportunity to present our latest work. Uh and also uh I would like to thank Otto and Lakshman to uh host us today. Also uh thanks I also would like to thank everyone who is here today with us. Um last time I fell ill and that's why we couldn't um uh we couldn't um make the webinar. Uh but hopefully it will be worth it for you to to wait. Uh so today I'll be presenting the concept of uh the tool that we recently developed to characterize dynamic protein interfaces and I'll explain what I mean by this and then after my presentation bin will have a demonstration of the tool uh so that the ones who would like to give it a try can have an opportunity to have an overview of how to run the tool on their own resources. Um so um we all know that we are living in very interesting times. Um currently uh modeling protein interactions um has been much easier compared to in the previous decades. um firstly because uh we have alpha faults and second we have a decades long accumulated knowledge on uh protein interaction prediction. So what was achieved by alpha for monomer modeling uh was also um reflected into multimer modeling. uh and currently especially if we talk about DRS as long as you have a good multiple sequence alignment for your partners uh it's very likely that you'll end up having a a good interaction model as an outcome of as an outcome of using alpha fold and different versions of alphault um which is something uh uh impressive if you are curious which tools which alpha fold uh tools to use for which type of uh uh modeling h you can try to mod you can try to follow the previous papers uh analysis community analysis papers published by Casper and Capri uh one of which is listed here uh as an outcome of this uh great advancement uh now we have uh we have so many models uh within our reach that are waiting for us uh to interpret them. Uh so for some time we had uh deposited models on monomers and their uh domains but just recently uh uh less than a month ago uh EBI extended their data sets by including several highquality homr and heterr structures. Um and actually it's not several but millions of structures. uh which means like actually now it's a good time uh to say okay uh when we have a confident prediction and we have many confident predictions now uh maybe it's time to move to a systematic analysis of how biological interfaces are formed and how are they are staying at the dynamics level and this is uh this sole aim of the project of the uh framework that we are presenting today uh we started Ed um the idea with the idea that um okay having interface analysis on static structures is is a great tool and it's very fast and we can do many things with it but now with the availability of so many models as well as availability of uh high high performance computing centers now we should also start doing um micro dynamic simulations of the interfaces I mean the whole systems in the end but focusing on the interfaces. There has been um a quite a number of uh efforts together MD data sets or having more like collective analysis on uh several uh um several MD systems. Uh but somehow the um analysis and the uh trajectory generation that are specifically focusing on the interface set has been missing and uh this has been on our list. So this gap has been on our to-do list already for a while especially since in um four papers already in the last couple of years. We have seen that if we simulate our systems and if we compare the interface dynamics at the atomistic level, we compare the types of interactions, if you compare the types of confirmational changes happening across the interfaces, we can actually very nicely and elegantly link um very tiny microscopic observations to the microscopic biological ones. So this uh we started uh with the idea started all in 2021 with a with a paralog system where three protein partners are getting together and they are structurally very similar but they had a different selectivity. So at first we we performed simulations on all systems and also on negative binders as well to understand what drives the selective binding and after this uh we uh actually extended the type of analysis and then the content and applied on other systems as well including protein nucleic acid systems or even systems as large as the nucleosome. So after having these observations we became very confident um that um actually um in terms of the analysis that we performed uh should be uh standardized should be released but it should be also shown that it can be applied in a general manner and for this we were very lucky because we encountered Sophia and Shantel from IBPC CNS Paris. Uh so they had also similar intentions. uh together uh with Bartin and me myself um as an outcome uh we decided to write a joint project. There was a call between uh France, a bilateral call between France and Turkey and then we got the project and as an outcome we um we formed a a beautiful collaboration uh where uh several MD data sets are generated by the France team and the analysis package and the analysis of the trajectories are performed by the Turkish team. uh so I'll briefly uh uh introduce Dynaben to you which is the benchmark dynamic benchmark that I that I just mentioned and then I'll discuss DAPIN in a bit more detail and then afterwards I will I'll leave the floor to Barin so that uh she will uh uh she can present the contents of our package. Okay. So at first Dina Dina bench uh this is actually the first systematic MD benchmark that is specifically focused on the interfaces. Of course what we do is we perform standard classical unbiased molecular dynamics. So there's nothing specific to interfaces at this point but we selected a data set that are specifically curated to understand different interactions and this is called docking benchmark 5.5. So earlier pre-alpha fault times meaning pre-AII times uh when someone was developing a new model to new um computational tool to predict protein interactions they always had to benchmark them on docking benchmark 5. So there was no training the methods were mostly physics based that's why it was totally okay but now together with the reduce of alpha fold as this benchmark is also a bit old um having this benchmark as a structure prediction benchmark became obsolete but it's still a very nicely curated and nicely classified data set because inside we have enzyme inhibitors antibi antigens they are all um functionally annotated they all have bound and unbound structures So it's a really nice resource to actually turn into an MD data set, MD benchmark. As an outcome, our French partners uh what they did was they took 200 complexes from this benchmark and simulated uh them each of them three times for 100 nconds. In an earlier work of theirs which actually which is the work that kind of united us in an earlier work of theirs they showed that um 100 nanconds uh standard should be okay to understand the interface dynamics. Okay, we know that it's a short time but here as I was mentioning our aim is to understand which interactions are stable, which interactions are specific, which interactions are transient and how is the interface staying consistent across 100 nconds. What they observed in their in a previous paper of theirs was uh even in the first 30 40 nanconds you observe if there's an interfacial change and that's why uh this time uh we hope should be enough. Uh so all our data sets and trajectories are online you can download them through uh this uh uh web link from this uh associated to uh CNRS. Um and basically you can also uh run use them for any type of analysis or training uh you want to have. Um and as the second part the second part is the part where we are mo mostly involved in together with Bin. Uh we are concentrated on developing and standardized analysis package. uh so one thing that was puzzling us when we were uh working in these previous biological examples that I was mentioning where we were performing comparative analysis to understand which interface behaves differently than others uh we used several different types of analysis and in the end by checking the literature as well we came up with a with a tool set analysis set that we think like should be standardly performed to understand the function of an interface uh and together with this we also have the uh we also have the uh aim to have uh to provide a standardized uh output for the users so that at least uh from this point and on MD results will be comparable to each other because as we all know the major problem in the MD analysis field is that everybody is using their own type of analysis which will help them to uncover the biological function they are researching. Um so I think for also developing any type of machine learning or AI approach we need the standardization and collectively gathering all these we hope that our dynamics uh dynamical dynamics analysis package which is focusing on characterizing interfaces in a standardized manner hopefully we'll find a nice place uh for itself in the community. Uh so coming to the architecture of our uh um of our tool uh that we called dynapin because basically we are analyzing the changing interactions across an interface over the time course of a trajectory. Um we in principle can get uh the input trajectory from all major uh simulation uh engines. Uh but you can also input uh PDB and sample. For example, you can also run uh NMR structures uh with this pipeline or you can mix models from different tools and then uh just feed them uh to the um to the analysis tool. Um the trajectory or the PDB should be accompanied with a topology file which could be a PSF or uh PDB. Uh there the program is going to read the chain ids etc all the properties. So it's important uh to remember of course that the topology and trajectory files should have matching atom content with matching chain content chain content and we have different types of configurations. If you don't touch anything by default it's the program is going to run everything that we are just going to describe but there's also an option to run everything individually as a separate analysis file analysis um command. Uh so the so in the core of the dynapin we have three modules uh which is quality control okay everyone runs quality control anyway so this is the most standardized part of MD analysis mentality but then we also offer two other uh modules uh one is residue based and the other one is interaction based and I'm going to come to them in the up in the uh upcoming slides and when the all the analysis are finished uh at the end we provide uh raw data in CSV format that is an outcome of uh all the calculations and we also generate uh automatically publication ready figures uh that are presenting the most interface-based um uh uh features that the user we hope would like to have. Um so when we since we are focused on understanding how the interfaces are changing over time, how the properties are changing over time, uh we first had to come up with a new interface definition for this type of analysis because um the first uh frame in your trajectory is different compared to the last and you of course will have different interface residues in both. So the first point for us to decide was okay first of all which interface definition we are going to use because there are different interface definitions we wanted someone to be fast but also accurate and that's why we resorted to the um famous uh definition uh from Emanuel Levi which is um a bit old right now but it's uh really nice and elegant. Uh so it's the it's basically classifies pines for the interface amino acids based on the solvent accessibility ch accessibility change of the residues upon complexation and then based on the solvent accessibility uh rate of change. It classifies the interface into three categories as support core and rim. Core is at the center super hydrophobic uh rim is at the outer ramps. For example, this is the place where you would find the salt bridges that are uh stabilizing the uh the edges of your interface and uh support is in between these two. uh and then uh what we also do is we calculate this interface residues over over the whole trajectory and we say by default if a residue is to be found at the interface for the 50% of the simulation time then this this amino acid is considered to be a part of the dynamic interface. Um the user can change uh this uh percentage it's not fixed. Uh but from our own internal test 50 turn out to be a good number. So after defining this dynamic uh interface throughout the whole trajectory or throughout your ensemble here is what we output at one go. So we have as quality control RMSD RMSF radius of generation and together with this we also have capri metrics. So for the people who are familiar to Capri, DOQ is a very well-known metric. But for the people who don't know this yet, DOQ is a metric that is composed of interfas, ligament MSD and fraction of native contacts and it actually measures how the complex changes from one state to the other and usually it's being used to assess uh protein interaction models but in this instance we use them to analyze the trajectory. Uh then as a second option we have this second module we have the residuebased module. Here we have all the relative solvent accessibilities of all amino acids. Uh we have the dynamic interfaces and their assignments either uh being at the core rim or the uh support. uh we have the secondary structure assignments so that for example the user can see whether there's any secondary structure change across the interface and we also nicely calculate electrostatics and wonder energies on a per residue basis. This was something that we were missing and we wanted to have this for our previous uh analysis in our previous projects. That's why we thought that it would be helpful to add this as well. And as a third module we have the non-covalent interaction module and we believe that this is the part um this would be maybe the most useful among all I mean all the uh options are useful but this we found to be most useful when we are uh trying to understand the impact of the interface on the function of the complex and here we provide at one go the uh intra interchain hydrogen bonds all at the atomistic level so bridges and hydrophobic context. Um so I'll just briefly um demonstrate a couple of analysis on a docking benchmark case that we have. Uh and then with this I'll I'll be finished with my part. Uh so the docking benchmark classifies the targets classifies the complexes according to their binding types as well as according to the confirmational change uh the monomers are um the monomers are experiencing upon binding. And here we demonstrate a case where the uh where the complex was characterized at difficult meaning like these two monomers are experiencing large scale confirmational change upon binding. And since this is a published structure, we also know we also could refer to its paper and from the paper we know that there is a very specific salt bridge uh which is helping the complex uh to get stabilized uh across its interface. So uh we run all the analysis at one go that I just mentioned and I'm only going to present to you uh a couple of analysis that we think should be useful for you to see. Um and here you see the output uh that we run to obtain all the analysis that I'm just going to mention. Uh so the file is coming from DAB and in Dynaben the trajectories are in the dry trajectories are in the DCD format. So that's why we are inputting a DCD file. uh and then we are getting out all analysis and all plots and we choose a resolution uh as 10 which is like analyze or pull uh every coordinate in every 10 steps. uh so what we observed for this complex for the duration of uh uh time scale was here we don't refer time uh on the x-axis because uh we know that uh people might be giving just a part of their trajectories or they might be giving a pdb ensemble that's why we uh always uh report things on a per frame basis so what we observed over here is that from the 200 frame and on the trajectory even from earlier on the trajectories is uh interface RMSD is very stable meaning like the interface is very stable as soon as you pass the equilibration the interface is behaving the same and here you see the uh root mean square fluctuations based on different chains chain A and chain B and the red dots represents the interface aminosis the dynamic uh interphase aminosis that are mapped on this plot and here you also see that the RMSFs of the interphase aminosis uh quite A number of them are below one extra and then a few of them are between one and two which also indicates uh which also indicates a rather stable uh interface. Uh then uh we pulled another uh figure uh which is the electrostatic energy um analysis. In this type of analysis, what we provide with is that uh we um um calculate per residue um electrostatic electrostatic energy throughout the whole frames and then we plot a box plot. We plot the distribution of the energies that we observed as an outcome of this. For example, what we can see that there is this lien uh 30 lizen 32. So there is a typo here. Sorry. Lizen 32 which is uh very significantly contributing to the inter have um across the surface and you can also see you can also see here that um uh we don't have many interface residues because these are all the uh inter we don't have many interface residues contributing to the electrostatics let's say uh because uh we have uh the plots only for the ones for which the electrostatics energy is calculated and this is also support bridges that we observe throughout the simulation and this kind of calculates our observ the observation that is coming from literature. So this is rather a proof of concept case in our um uh preprint. Uh we have uh more um deeper examples. I didn't want to present them here uh just to be able to concentrate on the package in instead of its biological applications. But we have better examples on our preprint if you are curious to have our preprint also raw data looks like. So these are all uh nicely organized CSV files uh that we are outputting. So if you don't like our figures uh we think that they are nice but you might not like them you can generate your own figures as well. So this is why we are providing the raw data as well also hopefully we are providing the raw data so that this type of data can be deposited together with the papers uh when uh people submit uh their MD trajectories uh to a journal. Okay with this uh my part is done. Uh so Bin is the one who is the who developed the analysis. Um so she is the uh uh main developer. We had support from uh other earlier lab members. Uh but in the end she is the one who made it uh so elegant and that's why I think I already talked enough. So Vin the floor is yours. Please um continue uh with the rest. >> Okay. Hello everyone again and thank you very much Esgie for that uh overview and the uh theoretical background of the motivation of the dynapin and now um I would like to uh take you all from the to the practice and firstly I would like to share my screen okay um first I will briefly walk through to the uh GitHub repository to show how Dynapin is uh structured, how it's installed and how it's used and then I will run a live example uh on my computer. Uh as you can see here on our repository page, we have a structured uh documentation to be as user friendly as possible. Uh let's quickly go through to the main sections. Um as uh as we mentioned, Dynapin provides uh through frame by frame uh analysis of molecular dynamics trajectories and it automatically extracts these dynamic interface metrics as we can see here. Uh now um in the K features uh I can say that Dynapin is organized around three main modules uh quality control, residue based and interaction based. So uh in the overall uh the workflow is uh built um three layers. First we assess the global behavior of the system. Then we characterize the interface uh residues um over time and finally we resolve these into the specific interaction patterns. Moving down to the system architecture. Dynapin is built as modular uh Python framework. We integrated commonly used uh tools such as MD analysis um PDB tools for trajectory handling uh free SUSA for uh um calculating uh solvent accessible surface area foldics for energetic analysis for secondary structure assignment and interphase for the interaction pro profiling uh within the same pipeline. Uh this means that uh users do not need to manage these components separately. Instead they can uh run a single workflow and obtain results uh in a standard design uh output structure with uh tables and figures. And in the system requirements section we summarize the basic setup uh needed to run the Dynapin. Uh, Dynapin supports um Linux and MacOss operating systems. Uh, Windows users also uh use Dynapin with VSSL. Uh, for installation uh users uh need Git uh cond to the Anaconda or Minionda and Python version 3.10 or higher. Um as a note uh here we specifically highlighted the foldex uh although uh all the packets used by dynamin installed your computer automatically with cond environment and foldex is an exception due to the due to the le licensing restrictions. If you want to run uh energetic analysis uh you need to download foldics separately and uh provide uh its path as a argument. And here uh you can see the installation uh section. The installation is intentionally simple. Uh first the repository is cloned, then the cond environment is created and finally it's activated. So with just few comments, Dynapin is ready to use. Um you just need to copy and paste these comments on your terminal. Uh once installed uh how do you use it? Um this section uh shows the main usage logic of Dynapin. We offer two main workflows. The first option is the command line interface. Um, Dynap can be run directly through the command line interface with a single comment. As you can see here, uh, these are the example commands for both PDB trajectories or PDB ensembles and a DCD trajectories uh with a topology file. And here is the other arguments. Uh just below the main uh command line arguments are listed. Uh very briefly these arguments uh control three things. The input files uh the analysis setup uh and the output location. Um I will also go through these com uh arguments in my demonstration. So uh this second option is the um interactive APA API workflow. Uh today I will show you in the demonstration part how you can see uh how you can use the uh provided Jupyter notebook uh and interactive API workflow. Yes. Uh here uh here uh in the output file section uh this is the output file organization. Uh the results are written into structured uh tables and figures uh directories and the outputs from the quality control residue based and interaction based analysis are clearly separated and systematically named here. Okay, then um I will now switch to the short demonstration with a Jupyter notebook. I uh have already activated my Dynapin cond environment and my uh uh let me show yes uh and my input trajectory and topology files are located in my working directory. This is the my um Jupyter notebook also. Uh this notebook is um exactly the same with the provided one uh in our GitHub GitHub repository. In the first step, I run this cell. Uh we simply import the core uh dynapin modules. Uh this step two is crucial because this is where we define uh all of our inputs uh parameters. Uh let me quickly go through the arguments. Uh we are setting here. Uh trajectory file and topology file. Uh these are our inputs. Uh today I am providing a DCD trajectory and my topology file here. Uh if you are using um uh PDB ensemble uh you can set this topology file to none. Um this job name dictates the name of the output directory where uh all our tables and figures will be safely stored. Uh this stride value for this demo uh I am intentionally using a relatively high uh stride value here. Uh let me increase it to 200. Um the reason is simply to keep the run short enough for this webinar instead of analyzing uh every single frame but you can keep this as one then you can analyze all of the frames in your um trajectory. Uh split models um argument is uh when we set this true dynapin uh will produce splitted models of your trajectory. Um I have set the chains uh argument to none uh to let Dynapin auto detect the interacting chains. But uh you could easily specify this something like um A and B. Uh dynopin uh allows to analyze dimer complexes. So if you have more than two chains in your complex, the chain selection allows uh you to restrict the analysis um to selected two chains. Uh threshold is the cutoff value uh cutff percentage of identifying dynamic interface residues as ESG mentioned in her presentation. Uh I leave it at the default 50%. Uh this means that if a residue stays at the interface layer um in other words support rim and core uh more than 50% of the simulation dynapin defines this residue uh as a dynamic interface residue and here lastly uh I defined here the foldics executable pet uh as I said uh if you don't want if you don't have the licensed fodex executable or do not wish to run energetic analysis. Uh you can set this none. Now I'm running this cell. Yes, in the step three uh before running the heavy calculations uh the notebook allows us to do a quick senate check using ngl wheel. I run this cell. Uh, and we can actually visualize our loaded trajectory here to ensure the um the chains look correct and your trajectory is uh okay. Then in this step four I initialize the dynapin uh engine. The setup has completed. Then now I am running the quality control analysis here. It's already completed. And here I run the residue based analysis. This is where Dynap calculates uh solvent accessible surface area um energetic calculations etc. And uh yes once that finished the notebook provides a really cool um interactive uh substep. Let me run this cell. Yes, we can visualize just the dynamic interface residues here, the ones that scored uh above our threshold value. Uh you can see the overall protein in transparent uh cartoon format while the dynamic interface residues are highlighted as a solid sticks. Moving to the step seven, the uh interaction based analysis that we execute This may take a while. Okay, this is completed. Uh now that all of our calculations are completed and the CSV files uh is written to the tables folder. uh as um as we already showed uh the overview of this table and so we initialize the plotter class here and I run this cell. The plots are generating. Here you can see a overview of the generated um plots. But uh as a last step uh we have a interactive uh result viewer and to visualize all the plots DAPIN generated like um interaction frequencies as a bar plot time evolution of the do Q score of native contacts and RMSD, RMSF and secondary structure uh assignments and the energy distribution of the um dynamic interface residues. Uh before I wrap up this demo, I want to quickly show you the um how Dynapin organized all these data. This is the um folder that Dynapin created and inside here we can see the two main subdirectories figures and tables. Uh in figures you can see all the high resolution um plots here like this. And the other uh directory is tables. Uh here you can uh also see the CSV files. If you don't um use to uh Detin figures and um if you want to do your own custom plotting or the statistical analysis later uh you can use these uh CSV files directly. Um I want to also point out these uh JSON files. These JSON files keep your um run parameters uh like this uh to reproduce your uh dynapin run. Again uh finally we also have some intermediate data. Uh this is the models uh zip file. Uh this uh keeps the um the the intermediate frames uh of your trajectory and the uh this fixed standard design PDB uh contains the uh standard design uh ensemble of your trajectory. Uh and that's it. uh from route trajectory to organized publication ready and uh ready data and visuals uh all in just in a few steps. Uh okay, that's it. Thank you for the for the demonstration. Yes, please auto. >> Yeah. Could I have the screen? Yes. So it's time for Q&A questions. We have two of them already here. So first one is thank you for the talk. This is great work. Fuzzy interfaces exist that do not respect the 50% threshold of contact time. Uh what do you recommend when using Dynapin for fuzzy interface analysis? Any caveats we should be wary of when running the code? Uh so I I can answer this first of all the nice words. Um all the interfaces uh that are being tested under the framework of um DAB as well as our previous interaction analysis the papers that I presented to you none of them contains uh contains a puzzy interface. uh so from my personal experience um I I can't tell a number uh but we've thought of such type of scenarios and considering this we change the cutoff we made the cutoff a variable so you can input your own cutoff uh but what I could advise to uh what I could advise for you would be to um run it with different uh thresholds and until you think like the threshold is small enough to produce understandable results because uh um depending on the uh threshold obviously the amount of interface residues that will be selected uh will differ. Uh so this is as much as I can say at the moment. I hope that this answers your question. >> Okay. Then next question is how do you analyze the residue residue decompos decomposition energies? Is it based on mmgbsa or mmmpsa? >> Uh so here uh let me answer this as well as bin mentioned h we use foldex. Uh the main reason why we use foldex is that uh it's uh it's a fast algorithm. It's very fast. So if we uh if we were to resort uh more uh accurate energy com calculations like MMPBSA uh then the calculation would have lasted quite long. Uh but since uh the trajectory is already chunked into uh pieces uh by the uh dynamin approach, one could also use uh these um uh frames um with other energy analysis tools if one has the resources for it. But here are choices based on the speed. >> Okay, there are no more questions. All right, there is actually one more right now. Uh, thank you for a nice demo. Do you plan to extend this to protein nucleic acid interfaces? Uh, and the second question regarding the MD benchmark data set on protein protein interfaces, do you know how how do AF generated ensembles compared to this benchmark? Thank you. >> Okay, thanks uh for these nice questions. So first of all maybe uh Bin claims that you can already use it for proteinic acid interactions right there's nothing against it right Barian >> yes yes but I couldn't test it enough for protein nucleic interactions >> but uh in the current version of the dynapin you should be able to uh calculate the proteinucle clear cast complexes. >> Mhm. So the we actually uh but how Bin framed the analysis package is that everything is written in a general way so that one could also use it for different types of molecules but we didn't want to we don't want to adverize it like the dynaben uh benchmark uh but in principle you can try to use it for uh protein nucleic acid interface phases and it would be great if you uh uh have a fe if you if you want to try and then you give us a feedback we would much appreciate this and regarding this second point so we actually have another project running uh for to answer your question uh but we are still at the preliminary stage so I cannot um answer your question at the moment but we are curious about this as well hopefully soon uh we'll have some preliminary data to present on the suspect too. >> Okay, no more questions it seems. Uh we still have time so we can if there's uh one more question popping up, we can get back to it quickly later. Meanwhile, I will take this opportunity to show us the upcoming webinars. Uh there will be one on 12th of May again driving forces in biomolecular condensates from atomistic simulations of model peptides. And then later in May on the 26th there will be con I need better glasses for this confir confirmational ensembles of intric intrically intrinsically disordered regions and proteins. Oh and uh one more shout out. There's a bioxal conference coming up in Berno in September and there's still it's still open for registration. Uh abstract deadline 1st of May. So I'll leave this up for a moment so you can scan the QR code if you're interested. Thank you everyone for joining us. >> Thanks a lot.