<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Gerard Martínez on Medium]]></title>
        <description><![CDATA[Stories by Gerard Martínez on Medium]]></description>
        <link>https://medium.com/@g.martinez_81343?source=rss-ee218b05d043------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*eMBX5Hum0rupjXeAVUHlSA.jpeg</url>
            <title>Stories by Gerard Martínez on Medium</title>
            <link>https://medium.com/@g.martinez_81343?source=rss-ee218b05d043------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 13:34:10 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@g.martinez_81343/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Using PlayMolecule® BindScope to participate in the SAMPL7 challenge [TUTORIAL]]]></title>
            <link>https://medium.com/playmolecule/using-playmolecule-bindscope-to-participate-in-the-sampl7-challenge-tutorial-2b24b9198dc2?source=rss-ee218b05d043------2</link>
            <guid isPermaLink="false">https://medium.com/p/2b24b9198dc2</guid>
            <category><![CDATA[structural-biology]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[drug-discovery]]></category>
            <category><![CDATA[protein]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Gerard Martínez]]></dc:creator>
            <pubDate>Tue, 03 Dec 2019 14:18:31 GMT</pubDate>
            <atom:updated>2019-12-05T10:05:15.098Z</atom:updated>
            <content:encoded><![CDATA[<h4>In this short tutorial we will look at how to use BindScope and, as an example, we will try to generate a submission for the first stage of the SAMPL7 challenge, in which the user is asked to discriminate between ligand binders and non-binders for a specific list of fragments against a protein</h4><h3>Two cents about the algorithm</h3><p>The algorithm behind BindScope is a 3D convolutional neural network designed as a binder/non-binder classifier. It has been trained using the <a href="http://dude.docking.org/">DUD-E</a> database (which contains ligands and decoys) and given a ligand docked to a protein, the algorithm is able to extract tridimensional structural features in the style of DeepSite and cast a prediction whether the ligand might be a binder. If you want to learn more about how the protein-ligand is featurized please check one of our old tutorials:</p><p><a href="https://medium.com/playmolecule/searching-for-binding-pockets-using-playmolecule-deepsite-tutorial-dc08316eb464">Searching for binding pockets using PlayMolecule® DeepSite [TUTORIAL]</a></p><p>As usual, for all the technical aspects please refer to the scientific <a href="https://academic.oup.com/bioinformatics/article-abstract/35/7/1237/5086393">publication</a>.</p><h3>The SAMPL7 challenge</h3><p>The first stage of the <a href="https://samplchallenges.github.io/">SAMPL7 challenge</a> consists in, given a list of 799 fragment SMILES and a protein structure(PHIP in this case2), finding which of these are actual binders. The organizers of the challenge, of course, have screened the whole library and know which are hits and which are not. You can find all the data available in the challenge github: <a href="https://github.com/samplchallenges/SAMPL7">https://github.com/samplchallenges/SAMPL7</a>.</p><h3>Applicability domain</h3><p>One problem we must highlight at this point is that BindScope may not work because the training database (DUDE) has little representation of fragments as shown in the plot below. This suggests that we might be outside of the applicability domain of BindScope because the training dataset differs substantially from the test dataset (the stage-1 SAMPL7 fragments). For the purpose of this tutorial we will try if it works anyway.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/627/1*hOzQCAGvcJ2RkPQkROkuhg.png" /><figcaption>Distribution of heavy atoms per ligand for the DUDE database (training dataset for BindScope) and the SAMPL7 fragments (our test set). Additionally we show the distribution for GDD (our GPCR-centric model). We can see how the distribution of SAMPL7 is clearly displaced towards the left and suggests we may be out of the applicability domain.</figcaption></figure><h3>Preparing the inputs</h3><p>BindScope expects tridimensional protein-ligand complexes so the first step is to generate binding poses for the list of fragment SMILES. For this we are going to use open source software, concretely we are going to generate starting 3D ligand conformations with <a href="https://www.rdkit.org/">RDkit</a> and use SMINA (which is based on <a href="http://vina.scripps.edu/manual.html">AutoDock VINA</a>) to generate docking poses. We will also need <a href="http://openbabel.org/wiki/Main_Page">Open Babel</a> on the way.</p><p>First, we are going to <strong>generate a single SDF </strong>with a starting conformation for each of the fragments:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/1a1b204bd73da2326465dd63e909ca8f/href">https://medium.com/media/1a1b204bd73da2326465dd63e909ca8f/href</a></iframe><p>Second, we are going to <strong>run SMINA</strong>. For this we will have to:</p><ol><li>Convert the original SAMPL7 protein PDB into the PDBQT format.</li></ol><pre>babel PHIPA_C2_Apo.pdb -xr -O PHIPA_C2_Apo.pdbqt</pre><p>2. Find out where is the center of the pocket. You can use your favorite molecular visualizer like VMD, pymol, etc. In the case of the challenge we can simply inspect the file called “PHIPA_C2_apo_sites.pdb” and we can find the binding sites reported by the challenge as the last 4 heteroatoms.</p><pre>HETATM 1290 HE S1 A1501 -19.150 12.842 24.700 0.44 10.58 HE<br>HETATM 1291 NE S2 A1601 -22.697 21.498 11.130 0.31 11.52 NE<br>HETATM 1289 AR S3 A1501 -6.393 4.395 15.539 0.52 26.08 AR<br>HETATM 1292 KR S4 A1501 -27.336 3.939 22.899 0.49 22.37 KR</pre><p>We will use only the first one for the purpose of this tutorial.</p><p>Finally with this information we run SMINA. We will ask it to generate 5 binding modes and assume a padding size of 15 Angstrom:</p><pre>smina.static  --receptor PHIPA_C2_Apo.pdbqt --ligand fragments.sdf --center_x -19.150 --center_y 12.842 --center_z 24.700 --size_x 15 --size_y 15 --size_z 15 --num_modes 5 --out docking.sdf</pre><p>The result is a set of 3045 docked fragments, which is a number slightly inferior to the asked 799 input fragments*5 binding modes (3995) but that’s because in some cases 5 modes couldn’t be generated and fewer were provided.</p><h3>Running BindScope</h3><p>With the previous docking SDF and the original PHIPA_C2_Apo.pdb we can now go to the BindScope application present in the PlayMolecule web platform: <a href="https://playmolecule.com/BindScope/">https://playmolecule.com/BindScope/</a>.</p><p>The input in this case is quite intuitive: simply upload the the PDB in the “Protein PDB” field and the ligands in the “Ligands” field. Since we already provide docked poses, make sure you leave the option “Perform ligand docking” unchecked. We will leave the remaining options by default and click Submit.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f2fwWsjulHY-RQqTzuJ_xw.png" /><figcaption>In case you want BindScope to dock your compounds you will have to place the yellow “search box” over the binding site of interest</figcaption></figure><p>Some curiosities about the app input in case you are wondering. Beware these options are not required for the SAMPL7 tutorial:</p><ol><li><strong>Model.</strong> While the default model was trained using the DUDE database, you can imagine that one could curate specific training datasets for specific types of proteins. For instance, we tried to include some extra GPCR protein-ligand binder/decoy complexes (<a href="http://cavasotto-lab.net/Databases/GDD/">GDD database</a>) and created a new model called BindScope_GPCR. In the model dropdown menu you will find BindScope_GPCR_transfer and BindScope_GPCR_scratch models: the first one was generated by fine tuning the original model with the new GPCR data for few epochs, while the second one was trained from scratch including a combination of DUDE data plus the extra GPCR data. If you are dealing with a GPCR system you might want to give these models a try.</li><li><strong>Docking</strong>. While the recommended use of the application is to provide already docked molecules to BindScope, if your molecules are not docked you can use the docking functionality. Simply click on the “Perform ligand docking” and move the yellow box around to place it in the pocket of interest. You can also modify the padding of the search box. As a warning: only the first 10 molecules will be docked so the free version of PlayMolecule is not an option if you are looking for high-throughput docking. This option is not required for the SAMPL7 tutorial because we are already providing docked poses.</li><li><strong>Stats</strong>. If your input SDF has some fields with integer 0s or 1s, it will be detected as possible “validation field”. This means you can provide the application with the “ground truth” (0 for non-binder, 1 for binder) and the application will run some automatic statistics about how good it was at predicting the labels. If the field selected by default does not correspond to a “validation field” then please set the “Ground truth” field to “None”.</li></ol><p>After running BindScope, the best ligands will be presented to you in the graphical user interface. As you can see, each ligand has an associated probability of binding. Only the first 100 ligands will be shown but if you want the whole list of results you can download it from the right panel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1DBTB1SQkiqC6XxVtWJJDg.png" /><figcaption>BindScope results interface</figcaption></figure><h3>Analyzing the results</h3><p>We now have for each fragment 1 to 5 binding scores provided by BindScope (depending on the number of binding modes SMINA was able to produce) and for each fragment we have a ground truth, which was made available at the challenge github: <a href="https://github.com/samplchallenges/SAMPL7/blob/master/protein_ligand/Stage-2-input-data/site-1_fragment-hits.csv">https://github.com/samplchallenges/SAMPL7/blob/master/protein_ligand/Stage-2-input-data/site-1_fragment-hits.csv</a> (In this CSV there’s a list of fragments that turned out to be hits, i.e. they bind the protein).</p><p>There are two ways (at least) we can proceed: one is to take the best BindScope score for each fragment, the other one is to take the mean of the scores provided by BindScope. If we sort our list of results by BindScope score and include the ground truth information, we can see that we detect at least two fragments in the top 15 predictions.</p><p>Here are the results taking the maxium BindScope score:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/4121b3afba0b7ed1661afb97c3dc19dc/href">https://medium.com/media/4121b3afba0b7ed1661afb97c3dc19dc/href</a></iframe><p>And here the results taking the mean BindScope score:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/953d8809fdea3942ceee632098da268b/href">https://medium.com/media/953d8809fdea3942ceee632098da268b/href</a></iframe><p>In both cases we are able to correctly discriminate fragments F275 and F309.</p><p>When looking at the AUC for both the methods we see it is not great: 0.5331 for the mean score and 0.5445 taking the max. BindScope score.</p><h3>Conclusion</h3><p>Even though the original BindScope model was trained with very few fragments present in the DUDE database, it was able to correctly discriminate 2 of the binding fragments in the top 3 (when using max. BindScope score) and 2 in the top 12 predictions when using the mean. However, no more hits were found in the near-top and the AUCs are not great. These results are not perfect but encouraging, and constitute a nice showcase of how to use BindScope in a real scenario. Results also show that BindScope holds potential for virtual screening and makes us wonder what would be the performance of a model trained exclusively on fragments for binding discrimination.</p><p>Thanks for reading!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/1*MLqaNc3_wCWBLodkPy8M2g.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2b24b9198dc2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/playmolecule/using-playmolecule-bindscope-to-participate-in-the-sampl7-challenge-tutorial-2b24b9198dc2">Using PlayMolecule® BindScope to participate in the SAMPL7 challenge [TUTORIAL]</a> was originally published in <a href="https://medium.com/playmolecule">PlayMolecule</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Searching for binding pockets using PlayMolecule® DeepSite [TUTORIAL]]]></title>
            <link>https://medium.com/playmolecule/searching-for-binding-pockets-using-playmolecule-deepsite-tutorial-dc08316eb464?source=rss-ee218b05d043------2</link>
            <guid isPermaLink="false">https://medium.com/p/dc08316eb464</guid>
            <category><![CDATA[drug-discovery]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[tutorial]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[structural-biology]]></category>
            <dc:creator><![CDATA[Gerard Martínez]]></dc:creator>
            <pubDate>Tue, 05 Nov 2019 11:47:21 GMT</pubDate>
            <atom:updated>2019-11-05T12:48:56.467Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rI05I_fR9a7w7st7emYQlQ.png" /></figure><h4>In this tutorial we will take a look at DeepSite, a web application to search for binding pockets using a 3D neural network-based predictor</h4><p>Welcome to another PlayMolecule tutorial. At this time, we will look at one of the first and most popular applications — DeepSite. While the algorithm is thoroughly explained in the corresponding <a href="https://www.ncbi.nlm.nih.gov/pubmed/28575181">paper</a>, in this article we will try to get an intuitive grasp and a practical approach to the application. Finally, we will demonstrate how to use the app by searching for pockets in the ABL Kinase (PDB code: 3CS9).</p><h4>The algorithm</h4><p>Deep learning has witnessed an amazing rate of development over the last few years leading to really mediatic successes in the field of machine vision. DeepSite leverages from the experience in the image recognition field and brings it to structural biology.</p><p>For deep learning to work, neural networks need to be trained with multiple examples (the more the merrier) of something we want to learn. In the classical “classification problem”, we want to present for instance, examples of two categories (e.g. photos of cats and photos that aren’t cats) and we want to the algorithm to be able to discern and correctly differentiate new examples of the category 1 from the category 2. DeepSite is not so different from this simple classifier, and what we do is that we take examples of binding pockets and non-binding pockets from the <a href="http://bioinfo-pharma.u-strasbg.fr/scPDB/">scPDB</a> database and we train a neural network to correctly classify what looks like a binding pocket from what doesn’t look like one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/305/1*9MMIYw19SLWaCulQGy7Sbw.png" /><figcaption><strong>Figure 1</strong>. DeepSite is not so different from a image-based cat classifier.</figcaption></figure><p>Since our protein structures are tridimensional and we want to use as much information as possible, we do not use a simple pictures of the pockets: we actually use a box that contains the pocket. In practice, this means that we use a type of neural networks called 3D convolutional neural networks, which are different from the ones used for image recognition (they use 2D convolution instead because pictures are 2-dimensional).</p><p>Further, in image recognition, the information about color is encoded and inputted into the network in the form of 3 channels (red, green, blue) that combined can reconstruct the original image.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/0*8fZChPlhteWXp2NK.jpg" /><figcaption><strong>Figure 2</strong>. In image recognition, a picture is decoded into three channels of color: red, green, blue</figcaption></figure><p>In DeepSite, we use the same principle to encode the information of the protein structure into several channels, each of them taking care of a different chemical property such as aromaticity, hydrophobicity, H-bond donor, H-bond acceptor, occupancy… Here’s a complete list of all the channels we generate:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/545/1*v9lNFoLwZZ7BpR2Ii_itCw.png" /><figcaption><strong>Figure 3</strong>. Channels generated by DeepSite to encode the structure into chemistry</figcaption></figure><p>And here is an example of how the hydrophobic and aromatic channels look for the PDBID 4NIE:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/546/1*rg39T0ulZ1Yyt8NC9vbdmQ.png" /><figcaption><strong>Figure 4</strong>. Hydrophobic (left) and aromatic (right) channels for the PDB ID 4NIE</figcaption></figure><p>This information is generated for 7622 proteins from the scPDB database for examples of boxes with “pocket” and “without pocket” and a 3D convolutional neural network (3D-CNN) is trained to differentiate these two classes.</p><p>Then, in inference time (i.e. when you submit your protein to the DeepSite application), we divide a protein into boxes and for each box we ask our pre-trained model: “is this a pocket?”. The model (3D-CNN) gives out a probability of being a pocket that ranges from 0 (not-pocket) to 1 (definitely-a-pocket). Finally, we aggregate this “probability in 3D space” into a so-called iso-surface, which is an interpolation of the different probabilities obtained from the per-box predictions. This isosurface helps us to highlight which area is more likely to be a pocket.</p><h4>An example</h4><p>DeepSite is very easy to use. To submit your protein, simply upload your PDB file or type a 4-letter RSCB PDB ID. In our case we will type “3CS9”, the ABL Kinase we are interested about. Finally we will select the chain “A”. The results will come up after a minute or so.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1002/1*niL2J9UlshrQ0-cswVi6qA.gif" /><figcaption><strong>Figure 5</strong>. DeepSite results for the PDB ID 3CS9.</figcaption></figure><p>As you can see, the generated isosurface is depicted in orange and represents the 3D volume where the probability of finding a binding pocket is the highest. You can change the cutoff of the isosurface by clicking on the dented wheel on the top right corner. Further, to simplify the analysis, the isosurface is clusterized into few clusters (putative pockets), the centers of which are shown in the right panel.</p><h4><strong>Conclusions</strong></h4><p>DeepSite leverages the technology developed for machine vision and image recognition and brings is to structural biology to detect pockets. Examples of boxes containing pockets and without pockets are extracted from proteins from the scPDB database and are used to train a 3D convolutional neural network. In prediction mode, your protein is broken into equally-sized boxes and for each box a prediction is casted. The different probabilities “in 3D space” are aggregated into isosurfaces, which are then shown to the user in the graphical user interface (GUI).</p><p>The DeepSite tool is freely available at <a href="http://www.playmolecule.com/">www.playmolecule.com</a> using public resources. For private installation, please get in touch with <a href="https://www.acellera.com/">Acellera</a> by contacting us at info@acellera.com.</p><p>Thanks for reading!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/1*MLqaNc3_wCWBLodkPy8M2g.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dc08316eb464" width="1" height="1" alt=""><hr><p><a href="https://medium.com/playmolecule/searching-for-binding-pockets-using-playmolecule-deepsite-tutorial-dc08316eb464">Searching for binding pockets using PlayMolecule® DeepSite [TUTORIAL]</a> was originally published in <a href="https://medium.com/playmolecule">PlayMolecule</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building and running a molecular dynamics (MD) simulation of a protein-ligand system on the…]]></title>
            <link>https://medium.com/playmolecule/building-and-running-a-molecular-dynamics-md-simulation-of-a-protein-ligand-system-on-the-a852ddb3eb87?source=rss-ee218b05d043------2</link>
            <guid isPermaLink="false">https://medium.com/p/a852ddb3eb87</guid>
            <category><![CDATA[tutorial]]></category>
            <category><![CDATA[molecular-dynamics]]></category>
            <category><![CDATA[structural-biology]]></category>
            <category><![CDATA[simulation]]></category>
            <category><![CDATA[web-applications]]></category>
            <dc:creator><![CDATA[Gerard Martínez]]></dc:creator>
            <pubDate>Mon, 04 Nov 2019 12:34:32 GMT</pubDate>
            <atom:updated>2019-11-04T12:34:32.313Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a-OPF08nGggXBYVJCG_Vpw.png" /></figure><h3>Building and running a molecular dynamics (MD) simulation of a protein-ligand system on the PlayMolecule® web platform [TUTORIAL]</h3><h4>In this article we will look at how to prepare, parameterize, build and run a molecular dynamics (MD) simulation of a protein-ligand system using the <a href="https://playmolecule.com/">PlayMolecule</a> platform</h4><p>Welcome to another PlayMolecule tutorial. In this occasion we will look at one of the most powerful features of <a href="https://playmolecule.com/">PlayMolecule</a>: our pipeline for building and running MD simulations. As you will see, the process is quite straightforward and thanks to PlayMolecule you should be able to do most of the actions required to run an MD simulation directly from your browser.</p><p>PlayMolecule consists of a set of applications that have the capability to “talk to each other” — this is, the output of one application can be used as input for another application. For instance, the SystemBuilder application requires a prepared protein structure that can be obtained through the application ProteinPrepare. Therefore, in order to build and run an MD simulation, we must use a number of applications in succession. Here’s an outline of the pipeline we will use in this tutorial:</p><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*XbxxT1r-npPO63jb-IiFEA.png" /><figcaption><strong>Figure 1</strong>. Pipeline to build and run an MD simulation</figcaption></figure><p>You can either run this tutorial from the public account (no login required) or in your personal account by logging in (top right corner) — in the latter case your structures will not be shown to the public. Either way, you must start and end the tutorial with same setup since the jobs from the public account are not visible from the private account and vice-versa.</p><p>As an example, in this tutorial we will build the well-known trypsin-benzamidine system. Without further ado, let’s learn how to do it.</p><h4>1. ProteinPrepare: preparing your protein structure</h4><p>The first step that we must follow is making our protein ready for MD. Luckily, that’s what the application <a href="https://playmolecule.com/proteinPrepare/">ProteinPrepare</a> was designed to do. We will not enter into the details of how to do it since it was the subject of another tutorial that you can visit below:</p><p><a href="https://medium.com/playmolecule/how-to-prepare-a-protein-structure-using-playmolecule-proteinprepare-tutorial-d69bd0a7c5b6">How to prepare a protein structure using PlayMolecule® ProteinPrepare [TUTORIAL]</a></p><p>Essentially, by using ProteinPrepare, we will input our initial structure and we will obtain a protonated and optimized structure that we will be able to use in SystemBuilder in a later step. To build our trypsin-benzamidine system, we will input the <em>3PTB</em> RSCB identifier into ProteinPrepare and we will specifically select to use the chain A.</p><h4>2. Parameterize: generate parameters for your ligand</h4><p>Parameterization is the process by which we can obtain force-field parameters that will describe how our ligand should behave in an MD simulation. To do so we can use the application <a href="https://playmolecule.com/parameterize/">Parameterize</a>, which will provide us AMBER-compatible parameters.</p><p>Parameterize is quite complex (you can read the full publication <a href="https://pubs.acs.org/doi/full/10.1021/acs.jcim.9b00439">here</a>) but in essence it takes care of:</p><ol><li>Computing the partial charge of the molecule atoms</li><li>Assigning general parameters (using the GAFF2 forcefield)</li><li>Optimizing dihedral parameters</li></ol><p>The third step is particularly important because general force-fields like GAFF2 are not really good at guessing the dihedral angle terms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*CbvysWxaLJZ-PlAa.png" /><figcaption><strong>Figure 2</strong>. Dihedral angle. Obtained from <a href="http://cbio.bmt.tue.nl/pumma/index.php/Theory/Potentials">http://cbio.bmt.tue.nl/pumma/index.php/Theory/Potentials</a></figcaption></figure><p>Parameterize can obtain the dihedral terms following one of the modes below:</p><ol><li>Simple <a href="http://ambermd.org/antechamber/gaff.html"><strong>GAFF2</strong></a> guessing (quick but inaccurate)</li><li>Dihedral scanning and parameter fitting using <a href="http://www.psicode.org/"><strong>QM calculations</strong></a> (slow but accurate)</li><li>Dihedral scanning and parameter fitting using the <a href="https://pubs.rsc.org/en/content/articlelanding/2017/sc/c6sc05720a#!divAbstract"><strong>ANI-1x</strong></a> neural network potential (a balance between speed and accuracy)</li></ol><p>You can select which mode you want to use from the graphical user interface. Beware that ANI1-x only accepts C,H,N,O atom types and both ANI-1x and QM modes have a max. limit of 25 atoms in the (free) public website.</p><p>Also bare in mind that while Parameterize takes care of generating parameters, it does not take care of “preparing” your ligand, this is it does not take care of its protonation. Actually, in Parameterize you can draw or input a ligand without hydrogens and they will be guessed using <a href="https://www.rdkit.org/">RDKit</a>, but its accuracy is dubious and we recommend using external software to come up with the right protonation state for your ligand.</p><p>Another remark: if you are planning to build a protein-ligand system with the ligand bound to the binding site, make sure the coordinates of the ligand are the correct (binding) ones. In this tutorial we assume you have the crystallographic or docking pose for your ligand.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pMrM440kv-LfWzzsSY6i_w.png" /><figcaption><strong>Figure 3.</strong> We download the SDF file from RSCB from PDB.</figcaption></figure><p>In our example, we will simply download and use the SDF of the benzamidine ligand directly from the 3PTB entry in RSCB: <a href="https://www.rcsb.org/structure/3PTB">https://www.rcsb.org/structure/3PTB</a> and rely on Parameterize for the protonation state. We will use the ANI-1x dihedral mode. Bare in mind these SDF files usually have multiple copies of the ligands for different chains but Parameterize will only use the first one, which in this case corresponds to the benzamidine that binds the chain A.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q8GaIupEhzlzUr0NPx3QaA.png" /><figcaption><strong>Figure 4</strong>. The parameters for Benzamidine obtained using Parameterize.</figcaption></figure><h4>3. SystemBuilder: building your protein-ligand system</h4><p>The next step in our pipeline is to build our system. Building a system means solvating (with water), ionizing and obtaining force-field parameters for the whole MD system. We can do so by using the <a href="https://playmolecule.com/SystemBuilder/">SystemBuilder</a> application, which is built on top of the Python framework <a href="https://www.acellera.com/products/high-throughput-molecular-dynamics/">HTMD</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*eede0k-ModCq843KGNFlGw.png" /><figcaption><strong>Figure 5</strong>. We select our protein and ligand from the drop-down menus.</figcaption></figure><p>In order to include the protein we previously prepared with ProteinPrepare and the ligand we just parameterized, we must select the jobs we previously executed from the drop down lists in the input form. For instance, in our case we select the protein <em>3ptb</em> and the ligand we identified as <em>Benzamidine</em>.</p><p>We leave the other options by default since we do not want to randomize the position of the ligand in the solvent (we want it to stay in the binding pocket as the initial coordinates provided) and we will leave the ionization to ‘0’, which means that ions will be added simply to neutralize the charge of the whole system to 0.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TleDxySNlmvVRyxv_njQmg.png" /><figcaption><strong>Figure 6</strong>. Output of SystemBuilder: our MD-ready protein-ligand system.</figcaption></figure><h4>4. SimpleRun: running a short MD simulation</h4><p>After building our system we can proceed to the <a href="https://playmolecule.com/SimpleRun/">SimpleRun</a> application, which uses our MD engine <a href="https://www.acellera.com/products/molecular-dynamics-software-gpu-acemd/">ACEMD</a> to run an MD simulation. The SimpleRun application essentially follows this three-step process:</p><ol><li>Minimization (forces are minimized, clashes are solved, etc.)</li><li>Equilibration run (MD is run for few nanoseconds with constraints so that the system doesn’t blow up)</li><li>Production run (MD without constraints)</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/336/1*RsP5Vw-BSgE2uwXayDDSSQ.png" /><figcaption><strong>Figure 7</strong>. Input options for SimpleRun.</figcaption></figure><p>To run SimpleRun, simply select the SystemBuilder job from the drop-down menu, select the length of the simulation and leave the rest of the options by default. Make sure that the ligand selected is the correct one and the constraints applied are “protein-ligand”. This is important because during equilibration we want the MD engine to apply constraints to our ligand as well (otherwise the ligand would be probably forced out of the pocket due to initial extreme forces).</p><p>Once you have run SimpleRun, you will be able to see an output like the one below. You can visualize the simulation by clicking on the “Play button”. You can also see some plots regarding the fluctuation of the ligand as well as a PDF report with all the information condensed, such as the previous plots, plus a protein-ligand interaction map (obtained with the <a href="https://playmolecule.com/PlexView/">PlexView</a> application) and other useful info. Additionally, three different types of clustering are applied to generate “representative” snapshots of your simulation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WzZind5uDEzqTFeazwoHEw.gif" /><figcaption><strong>Figure</strong> 8. SimpleRun output example for the 3UZA system.</figcaption></figure><h4>Conclusions</h4><p>The PlayMolecule platform is a suite of applications that can be used in succession to prepare, parameterize, build and run MD simulations — all from your browser. The public version of PlayMolecule platform is available for everyone to use at <a href="https://playmolecule.com/">https://playmolecule.com</a> and we hope it serves academics, students and industry to run MD simulations and learn about the dynamical properties of structural systems. The PlayMolecule platform is also available for local and private installation. If you are interested get in touch with us at info@acellera.com.</p><p>Thank you for reading!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/1*MLqaNc3_wCWBLodkPy8M2g.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a852ddb3eb87" width="1" height="1" alt=""><hr><p><a href="https://medium.com/playmolecule/building-and-running-a-molecular-dynamics-md-simulation-of-a-protein-ligand-system-on-the-a852ddb3eb87">Building and running a molecular dynamics (MD) simulation of a protein-ligand system on the…</a> was originally published in <a href="https://medium.com/playmolecule">PlayMolecule</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to prepare a protein structure using PlayMolecule® ProteinPrepare [TUTORIAL]]]></title>
            <link>https://medium.com/playmolecule/how-to-prepare-a-protein-structure-using-playmolecule-proteinprepare-tutorial-d69bd0a7c5b6?source=rss-ee218b05d043------2</link>
            <guid isPermaLink="false">https://medium.com/p/d69bd0a7c5b6</guid>
            <category><![CDATA[protein]]></category>
            <category><![CDATA[molecular-dynamics]]></category>
            <category><![CDATA[web-services]]></category>
            <category><![CDATA[drug-discovery]]></category>
            <category><![CDATA[structural-biology]]></category>
            <dc:creator><![CDATA[Gerard Martínez]]></dc:creator>
            <pubDate>Wed, 30 Oct 2019 12:22:50 GMT</pubDate>
            <atom:updated>2019-10-30T12:22:50.940Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F91wGgGqAZQrSebkgWhm0Q.png" /></figure><h4>In this short article we will look at how to use the <a href="https://playmolecule.com/proteinPrepare/">ProteinPrepare</a> application available at <a href="https://playmolecule.com/">PlayMolecule</a>, a platform for machine learning and molecular dynamics applications</h4><p>Welcome to PlayMolecule tutorials. In this series of articles we will explore some of the functionalities of the PlayMolecule platform. In this occasion we will take a look at ProteinPrepare, one of the most popular applications in our platform. Let’s dig right into it.</p><h4>Use case</h4><p>ProteinPrepare was designed to protonate and optimize a protein structure (PDB), a necessary step before building and running molecular dynamics applications. For this, we employ <a href="https://github.com/jensengroup/propka-3.1">PropKa</a> to predict the protonation state of the protein residues and <a href="https://www.ics.uci.edu/~dock/pdb2pqr/userguide.html">PDB2PQR</a> for the optimization of the hydrogen bond (H-bond) network. The expected output is a protein structure with the most likely protonation state (given a provided pH) and with an optimized hydrogen bond network. As usual, for all the technical details please take a look at our <a href="https://pubs.acs.org/doi/abs/10.1021/acs.jcim.7b00190">publication</a>.</p><h4>Application input</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/333/1*eSCr1AEEmO56jpoCRcrCcA.png" /><figcaption>ProteinPrepare input interface</figcaption></figure><p>The application expects the following inputs:</p><ol><li><strong>PDB id or PDB file</strong>. You can either provide an <a href="https://www.rcsb.org/">RSCB</a> PDB id (for instance <em>3PTB</em>) or your own PDB file (second field).</li><li><strong>pH</strong>. The pH at which you want to run your simulation and therefore protonate your protein residues. This is important because each protein aminoacid co-exists in different protonation states and whose proportion depends on the environment pH. For instance, a lower pH will increase the proportion of protonated aminoacids.</li><li><strong>Water removal</strong>. Water can establish meaningful H-bonds which can contribute to stabilize your protein structure. This option allows you to remove the water molecules present in the structure.</li><li><strong>Heteroatoms/ligands inclusion</strong>. While PropKa will not protonate ligands, cofactors or anything not-protein, their presence can be taken into account when running PropKa to determine the pKa (and therefore the protonation states) of the protein residues. Click the option to include the heteroatoms.</li><li><strong>Mutation</strong> (optionally). If your protein contains non-standard residues, you will be prompted with the possibility to mutate these to standard residues.</li></ol><h4>Application output</h4><p>The application usually runs smoothly and will give you the results upon seconds (depending on the queue availability). The output will look similar to this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aM_311fCSVFnCOtjv0h4ew.png" /><figcaption>ProteinPrepare output interface</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/219/1*tMDDf-AWOBZjQz2SecVkJg.png" /><figcaption>NGL viewer mouse commands</figcaption></figure><p>The central section will consist of the molecular visualizer <a href="http://nglviewer.org/ngl/api/">NGL viewer</a>. While its use is quite intuitive, here you can see the basic mouse commands to move the structure.</p><p>As you can see in the image, the protein backbone will become visible in the molecular viewer with few (or none) residues highlighted. These residues are also listed on the table at the right panel and correspond to aminoacids whose pKa is within 1 unit of the desired pH. You can visualize the hydrogen-bond network of each residue by clicking on it. We say that these residues have “dubious” protonation and require manual inspection because at the selected pH different protonation states co-exist with populations that range the 10–90% depending on how close the pKa is to the pH. This is a problem because in classical MD protein residues are fixed and cannot automatically change or flip their protonation states spontaneously (look at constant-pH simulations if you need so).</p><p>We can get more information if we look at the “Protonation diagram” tab (on the lower left corner).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LlQtHVlAel9FlSZfjJgzeA.png" /><figcaption>Protonation diagram</figcaption></figure><p>This diagram shows us the selected pH as a black thick vertical line and for each row, each aminoacid is depicted with its own pKa (white intersection), which can be described as the pH at which the residue co-exists in the protonated and de-protonated form (50% protonated and 50% de-protonated). Notice that the color <em>grey</em> denotes the “neutral form” of the residue, the color <em>red</em> means the residue is in its “negative form” and the color <em>blue</em> shows that the residue is in its “positive form”.</p><p>The vertical dashed lines at each side of the user-selected pH, depict +1 and -1 pH units around the pH. This is the “danger” zone. Residues that fall within this range will co-exist with different protonation states. For instance, Aspartic Acid (ASP) 52 (in the example above) has a pKa of 7.3 which lies slightly below the user-selected pH of 7.4. This means that the residue will co-exist in neutral and positive form at around 50% probability.</p><p>In the case of GLU 13, the situation is slightly less worrying since the pKa is further away from the pH, rendering the protonated form much more likely than the unprotonated form at around 80% protonated/20% unprotonated. In this case, the decision is easier and we might most likely want to simulate the protonated form.</p><p>PlayMolecule offers the “most likely” protonation form based on the provided pH and the micro-environment around each residue. However, there are cases in which you might want to change the protonation because, for instance, you might have the intuition that a specific hydrogen might be interacting/stabilizing your ligand of interest. In these cases, you can change the protonation state by using the “select” button (in the case of GLU or ASP) or the “dropdown” menu in the case of histidines. Both options are available at the right top panel. In the case of histidines, three options are available: HIP, HIE and HID. Their protonation differs like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/407/1*m3oa2M0YoezkyVHhXlOeXw.png" /><figcaption>Histidine in its different protonation states</figcaption></figure><p>Once you have made all the desired changes in protonation, then you can click “Re-prepare protein” and ProteinPrepare will be run again to apply the selected changes. Finally, you can download the protein by clicking the “Download current PDB” button.</p><h4>Conclusions</h4><p>PlayMolecule ProteinPrepare is a tool based on propKa and pdb2pqr that allows to guess the most suitable protonation states and optimize the H-bond network of your protein structure of interest. This is a necessary step beforing building and running molecular dynamics (MD) simulations. Its graphical user interface allows the user to override the proposed protonation states based on its own interests and intuition. The tool is freely available at <a href="http://www.playmolecule.com/">www.playmolecule.com</a> using public resources. For private installation, please get in touch with <a href="https://www.acellera.com/">Acellera</a> by contacting us at info@acellera.com.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/1*MLqaNc3_wCWBLodkPy8M2g.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d69bd0a7c5b6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/playmolecule/how-to-prepare-a-protein-structure-using-playmolecule-proteinprepare-tutorial-d69bd0a7c5b6">How to prepare a protein structure using PlayMolecule® ProteinPrepare [TUTORIAL]</a> was originally published in <a href="https://medium.com/playmolecule">PlayMolecule</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[PlayMolecule v2.0 release]]></title>
            <link>https://medium.com/playmolecule/playmolecule-v2-0-release-aa5f27b19ee4?source=rss-ee218b05d043------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa5f27b19ee4</guid>
            <category><![CDATA[neural-networks]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[chemistry]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[molecular-dynamics]]></category>
            <dc:creator><![CDATA[Gerard Martínez]]></dc:creator>
            <pubDate>Wed, 03 Oct 2018 10:34:10 GMT</pubDate>
            <atom:updated>2018-10-03T10:34:10.789Z</atom:updated>
            <content:encoded><![CDATA[<p>With the summer heat already waving goodbye and the batteries charged for a new exciting year, we are proud to announce the release of PlayMolecule v2.0! While only few changes are visible from the outside, as the popular saying states: “the beauty lies in the inside”.</p><h4><strong>What changes?</strong></h4><ul><li>We have undergone a modularity cleanup of all the applications, which are now highly reproducible and portable.</li><li>We have also released a new landing page, which now gives a clearer overview of the applications offered using a discovery pipeline chart as reference. Do you like the new <em>particles</em> background?</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FTukw3CoWiB2F7ed9SEHYA.png" /><figcaption>New landing page</figcaption></figure><ul><li>In-built benchmarking functionality. We’ve seen that sometimes very simple things can become a real hassle: benchmarking PlayMolecule algorithms can be one. For this reason, in the KDEEP and BindScope applications, if you provide the ground truth (pKd/ΔG for KDEEP and 0/1 for BindScope) we will be able to calculate how good are the algorithms at predicting it. If your input SDF includes a field with the correct type (float for KDEEP, 0/1 integers for BindScope) you will be able to select it as benchmark field, and we will automatically calculate AUC/ROC curve for BindScope and Pearson/Spearman correlation for KDEEP.</li></ul><h4>What else you’ve got?</h4><ul><li>As you may have noticed, we have also decided to start a new <a href="https://medium.com/playmolecule">blog</a> that we will use from now to communicate releases, news, small PlayMolecule tutorials or any information we may feel like sharing with you. We believe that PlayMolecule is nothing without a strong community willing to test and push the boundaries of what’s currently possible and that’s why we want to open this channel of communication with you.</li><li>We have also created a new Chat Room that you will find on the lower right corner of most of the PlayMolecule pages. This aligns with the previous point: we want to hear from you!</li><li>We have also made public a new app scoring functionality that aims to evaluate the experience of the users for each of the applications. We have been collecting your opinions for some weeks and we are now starting to share the first results. Thank you for your opinions!</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/373/1*LjP4KBZax3XdmNgFAQoMsw.png" /><figcaption>Each app now displays the average user rating</figcaption></figure><ul><li>We have also created a new user page that can be accessed through the <a href="http://www.playmolecule.org/user">/user</a> endpoint. For now, if you are logged in, you will be able to get an overview of the latest jobs you have executed.</li><li>We have seen that sometimes it is hard to keep track of the test datasets… In order to keep them organized and being able to use files bigger than 50 Mb, we have created a new page called <em>Data Center</em>. In the <em>Data Center</em> you can upload your files (for now only .<em>sdf</em> files), validate them and select them as input from inside the apps (for now available in BindScope and KDEEP). To upload files just drag &amp; drop them in the box below and click on “Upload files”.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VNnkzqGC8NU6ZFxZqVx7Ww.png" /><figcaption>Screenshot of the new section Data Center. Submit files, validate them and get them accessible from the apps.</figcaption></figure><p>That’s all for now folks! More content and news coming soon.</p><p><strong>PlayMolecule website:</strong> <a href="http://www.playmolecule.org/">http://www.playmolecule.org</a></p><p><strong>Twitter account:</strong> <a href="https://twitter.com/acellera?lang=es">https://twitter.com/acellera?lang=es</a></p><p><strong>Commercial solutions:</strong> <a href="https://www.acellera.com/">https://www.acellera.com</a></p><p><strong>Contact: </strong>info@acellera.com</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa5f27b19ee4" width="1" height="1" alt=""><hr><p><a href="https://medium.com/playmolecule/playmolecule-v2-0-release-aa5f27b19ee4">PlayMolecule v2.0 release</a> was originally published in <a href="https://medium.com/playmolecule">PlayMolecule</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>