Modeling conformational flexibility of kinases in inactive states.

Kinase structures in the inactive "DFG-out" state provide a wealth of druggable binding site variants. The conformational plasticity of this state can be mainly described by different conformations of binding site-forming elements such as DFG motif, A-loop, P-loop, and αC-helix. Compared to DFG-in structures, DFG-out structures are largely underrepresented in the Protein Data Bank (PDB). Thus, structure-based drug design efforts for DFG-out inhibitors may benefit from an efficient approach to generate an ensemble of DFG-out structures. Accordingly, the presented modeling pipeline systematically generates homology models of kinases in several DFG-out conformations based on a sophisticated creation of template structures that represent the major states of the flexible structural elements. Eighteen template classes were initially selected from all available kinase structures in the PDB and subsequently employed for modeling the entire kinome in different DFG-out variants by fusing individual structural elements to multiple chimeric template structures. Molecular dynamics simulations revealed that conformational transitions between the different DFG-out states generally do not occur within trajectories of a few hundred nanoseconds length. This underlines the benefits of the presented homology modeling pipeline to generate relevant conformations of "DFG-out" kinase structures for subsequent in silico screening or binding site analysis studies.

Mol2vec: Unsupervised Machine Learning Approach with Chemical Intuition.

Inspired by natural language processing techniques, we here introduce Mol2vec, which is an unsupervised machine learning approach to learn vector representations of molecular substructures. Like the Word2vec models, where vectors of closely related words are in close proximity in the vector space, Mol2vec learns vector representations of molecular substructures that point in similar directions for chemically related substructures. Compounds can finally be encoded as vectors by summing the vectors of the individual substructures and, for instance, be fed into supervised machine learning approaches to predict compound properties. The underlying substructure vector embeddings are obtained by training an unsupervised machine learning approach on a so-called corpus of compounds that consists of all available chemical matter. The resulting Mol2vec model is pretrained once, yields dense vector representations, and overcomes drawbacks of common compound feature representations such as sparseness and bit collisions. The prediction capabilities are demonstrated on several compound property and bioactivity data sets and compared with results obtained for Morgan fingerprints as a reference compound representation. Mol2vec can be easily combined with ProtVec, which employs the same Word2vec concept on protein sequences, resulting in a proteochemometric approach that is alignment-independent and thus can also be easily used for proteins with low sequence similarities.

KinMap: a web-based tool for interactive navigation through human kinome data.

BACKGROUND: Annotations of the phylogenetic tree of the human kinome is an intuitive way to visualize compound profiling data, structural features of kinases or functional relationships within this important class of proteins. The increasing volume and complexity of kinase-related data underlines the need for a tool that enables complex queries pertaining to kinase disease involvement and potential therapeutic uses of kinase inhibitors. RESULTS: Here, we present KinMap, a user-friendly online tool that facilitates the interactive navigation through kinase knowledge by linking biochemical, structural, and disease association data to the human kinome tree. To this end, preprocessed data from freely-available sources, such as ChEMBL, the Protein Data Bank, and the Center for Therapeutic Target Validation platform are integrated into KinMap and can easily be complemented by proprietary data. The value of KinMap will be exemplarily demonstrated for uncovering new therapeutic indications of known kinase inhibitors and for prioritizing kinases for drug development efforts. CONCLUSION: KinMap represents a new generation of kinome tree viewers which facilitates interactive exploration of the human kinome. KinMap enables generation of high-quality annotated images of the human kinome tree as well as exchange of kinome-related data in scientific communications. Furthermore, KinMap supports multiple input and output formats and recognizes alternative kinase names and links them to a unified naming scheme, which makes it a useful tool across different disciplines and applications. A web-service of KinMap is freely available at http://www.kinhub.org/kinmap/ .

Coupling Matched Molecular Pairs with Machine Learning for Virtual Compound Optimization.

Matched molecular pair (MMP) analyses are widely used in compound optimization projects to gain insights into structure-activity relationships (SAR). The analysis is traditionally done via statistical methods but can also be employed together with machine learning (ML) approaches to extrapolate to novel compounds. The here introduced MMP/ML method combines a fragment-based MMP implementation with different machine learning methods to obtain automated SAR decomposition and prediction. To test the prediction capabilities and model transferability, two different compound optimization scenarios were designed: (1) "new fragments" which occurs when exploring new fragments for a defined compound series and (2) "new static core and transformations" which resembles for instance the identification of a new compound series. Very good results were achieved by all employed machine learning methods especially for the new fragments case, but overall deep neural network models performed best, allowing reliable predictions also for the new static core and transformations scenario, where comprehensive SAR knowledge of the compound series is missing. Furthermore, we show that models trained on all available data have a higher generalizability compared to models trained on focused series and can extend beyond chemical space covered in the training data. Thus, coupling MMP with deep neural networks provides a promising approach to make high quality predictions on various data sets and in different compound optimization scenarios.

Complex long-distance effects of mutations that confer linezolid resistance in the large ribosomal subunit.

The emergence of multidrug-resistant pathogens will make current antibiotics ineffective. For linezolid, a member of the novel oxazolidinone class of antibiotics, 10 nucleotide mutations in the ribosome have been described conferring resistance. Hypotheses for how these mutations affect antibiotics binding have been derived based on comparative crystallographic studies. However, a detailed description at the atomistic level of how remote mutations exert long-distance effects has remained elusive. Here, we show that the G2032A-C2499A double mutation, located > 10 Å away from the antibiotic, confers linezolid resistance by a complex set of effects that percolate to the binding site. By molecular dynamics simulations and free energy calculations, we identify U2504 and C2452 as spearheads among binding site nucleotides that exert the most immediate effect on linezolid binding. Structural reorganizations within the ribosomal subunit due to the mutations are likely associated with mutually compensating changes in the effective energy. Furthermore, we suggest two main routes of information transfer from the mutation sites to U2504 and C2452. Between these, we observe cross-talk, which suggests that synergistic effects observed for the two mutations arise in an indirect manner. These results should be relevant for the development of oxazolidinone derivatives that are active against linezolid-resistant strains.

Identification and Visualization of Kinase-Specific Subpockets.

The identification and design of selective compounds is important for the reduction of unwanted side effects as well as for the development of tool compounds for target validation studies. This is, in particular, true for therapeutically important protein families that possess conserved folds and have numerous members such as kinases. To support the design of selective kinase inhibitors, we developed a novel approach that allows identification of specificity determining subpockets between closely related kinases solely based on their three-dimensional structures. To account for the intrinsic flexibility of the proteins, multiple X-ray structures of the target protein of interest as well as of unwanted off-target(s) are taken into account. The binding pockets of these protein structures are calculated and fused to a combined target and off-target pocket, respectively. Subsequently, shape differences between these two combined pockets are identified via fusion rules. The approach provides a user-friendly visualization of target-specific areas in a binding pocket which should be explored when designing selective compounds. Furthermore, the approach can be easily combined with in silico alanine mutation studies to identify selectivity determining residues. The potential impact of the approach is demonstrated in four retrospective experiments on closely related kinases, i.e., p38α vs Erk2, PAK1 vs PAK4, ITK vs AurA, and BRAF vs VEGFR2. Overall, the presented approach does not require any profiling data for training purposes, provides an intuitive visualization of a large number of protein structures at once, and could also be applied to other target classes.

Pocketome of human kinases: prioritizing the ATP binding sites of (yet) untapped protein kinases for drug discovery.

Protein kinases are involved in a variety of diseases including cancer, inflammation, and autoimmune disorders. Although the development of new kinase inhibitors is a major focus in pharmaceutical research, a large number of kinases remained so far unexplored in drug discovery projects. The selection and assessment of targets is an essential but challenging area. Today, a few thousands of experimentally determined kinase structures are available, covering about half of the human kinome. This large structural source allows guiding the target selection via structure-based druggability prediction approaches such as DoGSiteScorer. Here, a thorough analysis of the ATP pockets of the entire human kinome in the DFG-in state is presented in order to prioritize novel kinase structures for drug discovery projects. For this, all human kinase X-ray structures available in the PDB were collected, and homology models were generated for the missing part of the kinome. DoGSiteScorer was used to calculate geometrical and physicochemical properties of the ATP pockets and to predict the potential of each kinase to be druggable. The results indicate that about 75% of the kinome are in principle druggable. Top ranking structures comprise kinases that are primary targets of known approved drugs but additionally point to so far less explored kinases. The presented analysis provides new insights into the druggability of ATP binding pockets of the entire kinome. We anticipate this comprehensive druggability assessment of protein kinases to be helpful for the community to prioritize so far untapped kinases for drug discovery efforts.

Substrate derived peptidic α-ketoamides as inhibitors of the malarial protease PfSUB1.

Peptidic α-ketoamides have been developed as inhibitors of the malarial protease PfSUB1. The design of inhibitors was based on the best known endogenous PfSUB1 substrate sequence, leading to compounds with low micromolar to submicromolar inhibitory activity. SAR studies were performed indicating the requirement of an aspartate mimicking the P1' substituent and optimal P1-P4 length of the non-prime part. The importance of each of the P1-P4 amino acid side chains was investigated, revealing crucial interactions and size limitations.

Determinants of the species selectivity of oxazolidinone antibiotics targeting the large ribosomal subunit.

Oxazolidinone antibiotics bind to the highly conserved peptidyl transferase center in the ribosome. For developing selective antibiotics, a profound understanding of the selectivity determinants is required. We have performed for the first time technically challenging molecular dynamics simulations in combination with molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) free energy calculations of the oxazolidinones linezolid and radezolid bound to the large ribosomal subunits of the eubacterium Deinococcus radiodurans and the archaeon Haloarcula marismortui. A remarkably good agreement of the computed relative binding free energy with selectivity data available from experiment for linezolid is found. On an atomic level, the analyses reveal an intricate interplay of structural, energetic, and dynamic determinants of the species selectivity of oxazolidinone antibiotics: A structural decomposition of free energy components identifies influences that originate from first and second shell nucleotides of the binding sites and lead to (opposing) contributions from interaction energies, solvation, and entropic factors. These findings add another layer of complexity to the current knowledge on structure-activity relationships of oxazolidinones binding to the ribosome and suggest that selectivity analyses solely based on structural information and qualitative arguments on interactions may not reach far enough. The computational analyses presented here should be of sufficient accuracy to fill this gap.

Molecular determinants of binding to the Plasmodium subtilisin-like protease 1.

PfSUB1, a subtilisin-like protease of the human malaria parasite Plasmodium falciparum, is known to play important roles during the life cycle of the parasite and has emerged as a promising antimalarial drug target. In order to provide a detailed understanding of the origin of binding determinants of PfSUB1 substrates, we performed molecular dynamics simulations in combination with MM-GBSA free energy calculations using a homology model of PfSUB1 in complex with different substrate peptides. Key interactions, as well as residues that potentially make a major contribution to the binding free energy, are identified at the prime and nonprime side of the scissile bond and comprise peptide residues P4 to P2'. This finding stresses the requirement for peptide substrates to interact with both prime and nonprime side residues of the PfSUB1 binding site. Analyzing the energetic contributions of individual amino acids within the peptide-PfSUB1 complexes indicated that van der Waals interactions and the nonpolar part of solvation energy dictate the binding strength of the peptides and that the most favorable interactions are formed by peptide residues P4 and P1. Hot spot residues identified in PfSUB1 are dispersed over the entire binding site, but clustered areas of hot spots also exist and suggest that either the S4-S2 or the S1-S2' binding site should be exploited in efforts to design small molecule inhibitors. The results are discussed with respect to which binding determinants are specific to PfSUB1 and, therefore, might allow binding selectivity to be obtained.

Molecular Representations in Machine-Learning-Based Prediction of PK Parameters for Insulin Analogs.

Therapeutic peptides and proteins derived from either endogenous hormones, such as insulin, or de novo design via display technologies occupy a distinct pharmaceutical space in between small molecules and large proteins such as antibodies. Optimizing the pharmacokinetic (PK) profile of drug candidates is of high importance when it comes to prioritizing lead candidates, and machine-learning models can provide a relevant tool to accelerate the drug design process. Predicting PK parameters of proteins remains difficult due to the complex factors that influence PK properties; furthermore, the data sets are small compared to the variety of compounds in the protein space. This study describes a novel combination of molecular descriptors for proteins such as insulin analogs, where many contained chemical modifications, e.g., attached small molecules for protraction of the half-life. The underlying data set consisted of 640 structural diverse insulin analogs, of which around half had attached small molecules. Other analogs were conjugated to peptides, amino acid extensions, or fragment crystallizable regions. The PK parameters clearance (CL), half-life (T1/2), and mean residence time (MRT) could be predicted by using classical machine-learning models such as Random Forest (RF) and Artificial Neural Networks (ANN) with root-mean-square errors of CL of 0.60 and 0.68 (log units) and average fold errors of 2.5 and 2.9 for RF and ANN, respectively. Both random and temporal data splittings were employed to evaluate ideal and prospective model performance with the best models, regardless of data splitting, achieving a minimum of 70% of predictions within a twofold error. The tested molecular representations include (1) global physiochemical descriptors combined with descriptors encoding the amino acid composition of the insulin analogs, (2) physiochemical descriptors of the attached small molecule, (3) protein language model (evolutionary scale modeling) embedding of the amino acid sequence of the molecules, and (4) a natural language processing inspired embedding (mol2vec) of the attached small molecule. Encoding the attached small molecule via (2) or (4) significantly improved the predictions, while the benefit of using the protein language model-based encoding (3) depended on the used machine-learning model. The most important molecular descriptors were identified as descriptors related to the molecular size of both the protein and protraction part using Shapley additive explanations values. Overall, the results show that combining representations of proteins and small molecules was key for PK predictions of insulin analogs.

Development of Cagrilintide, a Long-Acting Amylin Analogue.

A hallmark of the pancreatic hormone amylin is its high propensity toward the formation of amyloid fibrils, which makes it a challenging drug design effort. The amylin analogue pramlintide is commercially available for diabetes treatment as an adjunct to insulin therapy but requires three daily injections due to its short half-life. We report here the development of the stable, lipidated long-acting amylin analogue cagrilintide (23) and some of the structure-activity efforts that led to the selection of this analogue for clinical development with obesity as an indication. Cagrilintide is currently in clinical trial and has induced significant weight loss when dosed alone or in combination with the GLP-1 analogue semaglutide.

Molecular recognition of RNA: challenges for modelling interactions and plasticity.

There is growing interest in molecular recognition processes of RNA because of RNA's widespread involvement in biological processes. Computational approaches are increasingly used for analysing and predicting binding to RNA, fuelled by encouraging progress in developing simulation, free energy and docking methods for nucleic acids. These developments take into account challenges regarding the energetics of RNA-ligand binding, RNA plasticity, and the presence of water molecules and ions in the binding interface. Accordingly, we will detail advances in force field and scoring function development for molecular dynamics (MD) simulations, free energy computations and docking calculations of nucleic acid complexes. Furthermore, we present methods that can detect moving parts within RNA structures based on graph-theoretical approaches or normal mode analysis (NMA). As an example of the successful use of these developments, we will discuss recent structure-based drug design approaches that focus on the bacterial ribosomal A-site RNA as a drug target.

HIV-1 TAR RNA spontaneously undergoes relevant apo-to-holo conformational transitions in molecular dynamics and constrained geometrical simulations.

We report all-atom molecular dynamics and replica exchange molecular dynamics simulations on the unbound human immunodeficiency virus type-1 (HIV-1) transactivation responsive region (TAR) RNA structure and three TAR RNA structures in bound conformations of, in total, approximately 250 ns length. We compare the extent of observed conformational sampling with that of the conceptually simpler and computationally much cheaper constrained geometrical simulation approach framework rigidity optimized dynamic algorithm (FRODA). Atomic fluctuations obtained by replica-exchange molecular dynamics (REMD) simulations agree quantitatively with those obtained by molecular dynamics (MD) and FRODA simulations for the unbound TAR structure. Regarding the stereochemical quality of the generated conformations, backbone torsion angles and puckering modes of the sugar-phosphate backbone were reproduced equally well by MD and REMD simulations, but further improvement is needed in the case of FRODA simulations. Essential dynamics analysis reveals that all three simulation approaches show a tendency to sample bound conformations when starting from the unbound TAR structure, with MD and REMD simulations being superior with respect to FRODA. These results are consistent with the experimental view that bound TAR RNA conformations are transiently sampled in the free ensemble, following a conformation selection model. The simulation-generated TAR RNA conformations have been successfully used as receptor structures for docking. This finding has important implications for RNA-ligand docking in that docking into an ensemble of simulation-generated RNA structures is shown to be a valuable means to cope with large apo-to-holo conformational transitions of the receptor structure.

Constraint counting on RNA structures: linking flexibility and function.

RNA structures are highly flexible biomolecules that can undergo dramatic conformational changes required to fulfill their diverse functional roles. Constraint counting on a topological network representation of an RNA structure can provide very efficiently detailed insights into the intrinsic flexibility characteristics of the biomolecule. In the network, vertices represent atoms and edges represent covalent and strong non-covalent bonds and angle constraints. Initially, the method has been successfully applied to identify rigid and flexible regions in proteins. Here, we present recent progress in extending the approach to RNA structures. As a case study, we analyze stability characteristics of the ribosomal exit tunnel and relate these findings to the tunnel's active role in co-translational processes.

Analyzing the flexibility of RNA structures by constraint counting.

RNA requires conformational dynamics to undergo its diverse functional roles. Here, a new topological network representation of RNA structures is presented that allows analyzing RNA flexibility/rigidity based on constraint counting. The method extends the FIRST approach, which identifies flexible and rigid regions in atomic detail in a single, static, three-dimensional molecular framework. Initially, the network rigidity of a canonical A-form RNA is analyzed by counting on constraints of network elements of increasing size. These considerations demonstrate that it is the inclusion of hydrophobic contacts into the RNA topological network that is crucial for an accurate flexibility prediction. The counting also explains why a protein-based parameterization results in overly rigid RNA structures. The new network representation is then validated on a tRNA(ASP) structure and all NMR-derived ensembles of RNA structures currently available in the Protein Data Bank (with chain length >/=40). The flexibility predictions demonstrate good agreement with experimental mobility data, and the results are superior compared to predictions based on two previously used network representations. Encouragingly, this holds for flexibility predictions as well as mobility predictions obtained by constrained geometric simulations on these networks. Potential applications of the approach to analyzing the flexibility of DNA and RNA/protein complexes are discussed.

Kinome-Wide Profiling Prediction of Small Molecules.

Extensive kinase profiling data, covering more than half of the human kinome, are available nowadays and allow the construction of activity prediction models of high practical utility. Proteochemometric (PCM) approaches use compound and protein descriptors, which enables the extrapolation of bioactivity values to thus far unexplored kinases. In this study, the potential of PCM to make large-scale predictions on the entire kinome is explored, considering the applicability on novel compounds and kinases, including clinically relevant mutants. A rigorous validation indicates high predictive power on left-out kinases and superiority over individual kinase QSAR models for new compounds. Furthermore, external validation on clinically relevant mutant kinases reveals an excellent predictive power for mutations spread across the ATP binding site.

From Cancer to Pain Target by Automated Selectivity Inversion of a Clinical Candidate.

Elimination of inadvertent binding is crucial for inhibitor design targeting conserved protein classes like kinases. Compounds in clinical trials provide a rich source for initiating drug design efforts by exploiting such secondary binding events. Considering both aspects, we shifted the selectivity of tozasertib, originally developed against AurA as cancer target, toward the pain target TrkA. First, selectivity-determining features in binding pockets were identified by fusing interaction grids of several key and off-target conformations. A focused library was subsequently created and prioritized using a multiobjective selection scheme that filters for selective and highly active compounds based on orthogonal methods grounded in computational chemistry and machine learning. Eighteen high-ranking compounds were synthesized and experimentally tested. The top-ranked compound has 10000-fold improved selectivity versus AurA, nanomolar cellular activity, and is highly selective in a kinase panel. This was achieved in a single round of automated in silico optimization, highlighting the power of recent advances in computer-aided drug design to automate design and selection processes.

Profiling Prediction of Kinase Inhibitors: Toward the Virtual Assay.

Kinome-wide screening would have the advantage of providing structure-activity relationships against hundreds of targets simultaneously. Here, we report the generation of ligand-based activity prediction models for over 280 kinases by employing Machine Learning methods on an extensive data set of proprietary bioactivity data combined with open data. High quality (AUC > 0.7) was achieved for ∼200 kinases by (1) combining open with proprietary data, (2) choosing Random Forest over alternative tested Machine Learning methods, and (3) balancing the training data sets. Tests on left-out and external data indicate a high value for virtual screening projects. Importantly, the derived models are evenly distributed across the kinome tree, allowing reliable profiling prediction for all kinase branches. The prediction quality was further improved by employing experimental bioactivity fingerprints of a small kinase subset. Overall, the generated models can support various hit identification tasks, including virtual screening, compound repurposing, and the detection of potential off-targets.

A validated antibody panel for the characterization of tau post-translational modifications.

BACKGROUND: Tau is a microtubule-binding protein, which is subject to various post-translational modifications (PTMs) including phosphorylation, methylation, acetylation, glycosylation, nitration, sumoylation and truncation. Aberrant PTMs such as hyperphosphorylation result in tau aggregation and the formation of neurofibrillary tangles, which are a hallmark of Alzheimer's disease (AD). In order to study the importance of PTMs on tau function, antibodies raised against specific modification sites are widely used. However, quality control of these antibodies is lacking and their specificity for particular modifications is often unclear. METHODS: In this study, we first designed an online tool called 'TauPTM', which enables the visualization of PTMs and their interactions on human tau. Using TauPTM, we next searched for commercially available antibodies against tau PTMs and characterized their specificity by peptide array, immunoblotting, electrochemiluminescence ELISA and immunofluorescence technologies. RESULTS: We demonstrate that commercially available antibodies can show a significant lack of specificity, and PTM-specific antibodies in particular often recognize non-modified versions of the protein. In addition, detection may be hindered by other PTMs in close vicinity, complicating the interpretation of results. Finally, we compiled a panel of specific antibodies and show that they are useful to detect PTM-modified endogenous tau in hiPSC-derived neurons and mouse brains. CONCLUSION: This study has created a platform to reliably and robustly detect changes in localization and abundance of post-translationally modified tau in health and disease. A web-based version of TauPTM is fully available at http://www.tauptm.org .