Predicting Protein-Peptide Interactions: Benchmarking Deep Learning Techniques and a Comparison with Focused Docking.

The accurate prediction of protein structures achieved by deep learning (DL) methods is a significant milestone and has deeply impacted structural biology. Shortly after its release, AlphaFold2 has been evaluated for predicting protein-peptide interactions and shown to significantly outperform RoseTTAfold as well as a conventional blind docking method: PIPER-FlexPepDock. Since then, new AlphaFold2 models, trained specifically to predict multimeric assemblies, have been released and a new ab initio folding model OmegaFold has become available. Here, we assess docking success rates for these new DL folding models and compare their performance with our state-of-the-art, focused peptide-docking software AutoDock CrankPep (ADCP). The evaluation is done using the same dataset and performance metric for all methods. We show that, for a set of 99 nonredundant protein-peptide complexes, the new AlphaFold2 model outperforms other Deep Learning approaches and achieves remarkable docking success rates for peptides. While the docking success rate of ADCP is more modest when considering the top-ranking solution only, it samples correct solutions for around 62% of the complexes. Interestingly, different methods succeed on different complexes, and we describe a consensus docking approach using ADCP and AlphaFold2, which achieves a remarkable 60% for the top-ranking results and 66% for the top 5 results for this set of 99 protein-peptide complexes.

Improving Docking Power for Short Peptides Using Random Forest.

In recent years, therapeutic peptides have gained a lot interest as demonstrated by the 60 peptides approved as drugs in major markets and 150+ peptides currently in clinical trials. However, while small molecule docking is routinely used in rational drug design efforts, docking peptides has proven challenging partly because docking scoring functions, developed and calibrated for small molecules, perform poorly for these molecules. Here, we present random forest classifiers trained to discriminate correctly docked peptides. We show that, for a testing set of 47 protein-peptide complexes, structurally dissimilar from the training set and previously used to benchmark AutoDock Vina's ability to dock short peptides, these random forest classifiers improve docking power from ∼25% for AutoDock scoring functions to an average of ∼70%. These results pave the way for peptide-docking success rates comparable to those of small molecule docking. To develop these classifiers, we compiled the ProptPep37_2021 data set, a curated, high-quality set of 322 crystallographic protein-peptides complexes annotated with structural similarity information. The data set also provides a collection of high-quality putative poses with a range of deviations from the crystallographic pose, providing correct and incorrect poses (i.e., decoys) of the peptide for each entry. The ProptPep37_2021 data set as well as the classifiers presented here are freely available.

Cyclic Peptides as Protein Kinase Inhibitors: Structure-Activity Relationship and Molecular Modeling.

Under-expression or overexpression of protein kinases has been shown to be associated with unregulated cell signal transduction in cancer cells. Therefore, there is major interest in designing protein kinase inhibitors as anticancer agents. We have previously reported [WR]5, a peptide containing alternative arginine (R) and tryptophan (W) residues as a non-competitive c-Src tyrosine kinase inhibitor. A number of larger cyclic peptides containing alternative hydrophobic and positively charged residues [WR]x (x = 6-9) and hybrid cyclic-linear peptides, [R6K]W6 and [R5K]W7, containing R and W residues were evaluated for their protein kinase inhibitory potency. Among all the peptides, cyclic peptide [WR]9 was found to be the most potent tyrosine kinase inhibitor. [WR]9 showed higher inhibitory activity (IC50 = 0.21 µM) than [WR]5, [WR]6, [WR]7, and [WR]8 with IC50 values of 0.81, 0.57, 0.35, and 0.33 µM, respectively, against c-Src kinase as determined by a radioactive assay using [γ-33P]ATP. Consistent with the result above, [WR]9 inhibited other protein kinases such as Abl kinase activity with an IC50 value of 0.35 µM, showing 2.2-fold higher inhibition than [WR]5 (IC50 = 0.79 µM). [WR]9 also inhibited PKCa kinase activity with an IC50 value of 2.86 µM, approximately threefold higher inhibition than [WR]5 (IC50 = 8.52 µM). A similar pattern was observed against Braf, c-Src, Cdk2/cyclin A1, and Lck. [WR]9 exhibited IC50 values of <0.25 µM against Akt1, Alk, and Btk. These data suggest that [WR]9 is consistently more potent than other cyclic peptides with a smaller ring size and hybrid cyclic-linear peptides [R6K]W6 and [R5K]W7 against selected protein kinases. Thus, the presence of R and W residues in the ring, ring size, and the number of amino acids in the structure of the cyclic peptide were found to be critical in protein kinase inhibitory potency. We identified three putative binding pockets through automated blind docking of cyclic peptides [WR](5-9). The most populated pocket is located between the SH2, SH3, and N-lobe domains on the opposite side of the ATP binding site. The second putative pocket is formed by the same domains and located on the ATP binding site side of the protein. Finally, a third pocket was identified between the SH2 and SH3 domains. These results are consistent with the non-competitive nature of the inhibition displayed by these molecules. Molecular dynamics simulations of the protein-peptide complexes indicate that the presence of either [WR]5 or [WR]9 affects the plasticity of the protein and in particular the volume of the ATP binding site pocket in different ways. These results suggest that the second pocket is most likely the site where these peptides bind and offer a plausible rationale for the increased affinity of [WR]9.

AutoDock CrankPep: combining folding and docking to predict protein-peptide complexes.

MOTIVATION: Protein-peptide interactions mediate a wide variety of cellular and biological functions. Methods for predicting these interactions have garnered a lot of interest over the past few years, as witnessed by the rapidly growing number of peptide-based therapeutic molecules currently in clinical trials. The size and flexibility of peptides has shown to be challenging for existing automated docking software programs. RESULTS: Here we present AutoDock CrankPep or ADCP in short, a novel approach to dock flexible peptides into rigid receptors. ADCP folds a peptide in the potential field created by the protein to predict the protein-peptide complex. We show that it outperforms leading peptide docking methods on two protein-peptide datasets commonly used for benchmarking docking methods: LEADS-PEP and peptiDB, comprised of peptides with up to 15 amino acids in length. Beyond these datasets, ADCP reliably docked a set of protein-peptide complexes containing peptides ranging in lengths from 16 to 20 amino acids. The robust performance of ADCP on these longer peptides enables accurate modeling of peptide-mediated protein-protein interactions and interactions with disordered proteins. AVAILABILITY AND IMPLEMENTATION: ADCP is distributed under the LGPL 2.0 open source license and is available at http://adcp.scripps.edu. The source code is available at https://github.com/ccsb-scripps/ADCP. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

AutoGridFR: Improvements on AutoDock Affinity Maps and Associated Software Tools.

Precomputed affinity maps are used by AutoDock to efficiently describe rigid biomolecules called receptors in automated docking. These maps greatly speed up the docking process and allow users to experiment with the forcefield. Here, we present AutoGridFR (AGFR): a software tool facilitating the calculation of these maps. We describe a new version of the AutoSite algorithm that improves the description of binding pockets automatically detected on receptors, and an algorithm for adding affinity gradients which help search methods optimize solution using fewer evaluations of the scoring functions. AGFR supports the calculation of maps for various advanced docking techniques such as covalent docking, hydrated docking, and docking with flexible receptor sidechains. Maps are stored in a single file along with metadata supporting data provenance, reproducibility, and facilitating their management. Finally, maps can be calculated from the command line or through a modern graphical user interface which also supports their visualization. © 2019 Wiley Periodicals, Inc.

cellPACK: a virtual mesoscope to model and visualize structural systems biology.

cellPACK assembles computational models of the biological mesoscale, an intermediate scale (10-100 nm) between molecular and cellular biology scales. cellPACK's modular architecture unites existing and novel packing algorithms to generate, visualize and analyze comprehensive three-dimensional models of complex biological environments that integrate data from multiple experimental systems biology and structural biology sources. cellPACK is available as open-source code, with tools for validation of models and with 'recipes' and models for five biological systems: blood plasma, cytoplasm, synaptic vesicles, HIV and a mycoplasma cell. We have applied cellPACK to model distributions of HIV envelope protein to test several hypotheses for consistency with experimental observations. Biologists, educators and outreach specialists can interact with cellPACK models, develop new recipes and perform packing experiments through scripting and graphical user interfaces at http://cellPACK.org/.

AutoSite: an automated approach for pseudo-ligands prediction-from ligand-binding sites identification to predicting key ligand atoms.

MOTIVATION: The identification of ligand-binding sites from a protein structure facilitates computational drug design and optimization, and protein function assignment. We introduce AutoSite: an efficient software tool for identifying ligand-binding sites and predicting pseudo ligand corresponding to each binding site identified. Binding sites are reported as clusters of 3D points called fills in which every point is labelled as hydrophobic or as hydrogen bond donor or acceptor. From these fills AutoSite derives feature points: a set of putative positions of hydrophobic-, and hydrogen-bond forming ligand atoms. RESULTS: We show that AutoSite identifies ligand-binding sites with higher accuracy than other leading methods, and produces fills that better matches the ligand shape and properties, than the fills obtained with a software program with similar capabilities, AutoLigand In addition, we demonstrate that for the Astex Diverse Set, the feature points identify 79% of hydrophobic ligand atoms, and 81% and 62% of the hydrogen acceptor and donor hydrogen ligand atoms interacting with the receptor, and predict 81.2% of water molecules mediating interactions between ligand and receptor. Finally, we illustrate potential uses of the predicted feature points in the context of lead optimization in drug discovery projects. AVAILABILITY AND IMPLEMENTATION: http://adfr.scripps.edu/AutoDockFR/autosite.html CONTACT: sanner@scripps.eduSupplementary information: Supplementary data are available at Bioinformatics online.

AutoDockFR: Advances in Protein-Ligand Docking with Explicitly Specified Binding Site Flexibility.

Automated docking of drug-like molecules into receptors is an essential tool in structure-based drug design. While modeling receptor flexibility is important for correctly predicting ligand binding, it still remains challenging. This work focuses on an approach in which receptor flexibility is modeled by explicitly specifying a set of receptor side-chains a-priori. The challenges of this approach include the: 1) exponential growth of the search space, demanding more efficient search methods; and 2) increased number of false positives, calling for scoring functions tailored for flexible receptor docking. We present AutoDockFR-AutoDock for Flexible Receptors (ADFR), a new docking engine based on the AutoDock4 scoring function, which addresses the aforementioned challenges with a new Genetic Algorithm (GA) and customized scoring function. We validate ADFR using the Astex Diverse Set, demonstrating an increase in efficiency and reliability of its GA over the one implemented in AutoDock4. We demonstrate greatly increased success rates when cross-docking ligands into apo receptors that require side-chain conformational changes for ligand binding. These cross-docking experiments are based on two datasets: 1) SEQ17 -a receptor diversity set containing 17 pairs of apo-holo structures; and 2) CDK2 -a ligand diversity set composed of one CDK2 apo structure and 52 known bound inhibitors. We show that, when cross-docking ligands into the apo conformation of the receptors with up to 14 flexible side-chains, ADFR reports more correctly cross-docked ligands than AutoDock Vina on both datasets with solutions found for 70.6% vs. 35.3% systems on SEQ17, and 76.9% vs. 61.5% on CDK2. ADFR also outperforms AutoDock Vina in number of top ranking solutions on both datasets. Furthermore, we show that correctly docked CDK2 complexes re-create on average 79.8% of all pairwise atomic interactions between the ligand and moving receptor atoms in the holo complexes. Finally, we show that down-weighting the receptor internal energy improves the ranking of correctly docked poses and that runtime for AutoDockFR scales linearly when side-chain flexibility is added.

The AutoDock suite at 30.

The AutoDock suite provides a comprehensive toolset for computational ligand docking and drug design and development. The suite builds on 30 years of methods development, including empirical free energy force fields, docking engines, methods for site prediction, and interactive tools for visualization and analysis. Specialized tools are available for challenging systems, including covalent inhibitors, peptides, compounds with macrocycles, systems where ordered hydration plays a key role, and systems with substantial receptor flexibility. All methods in the AutoDock suite are freely available for use and reuse, which has engendered the continued growth of a diverse community of primary users and third-party developers.

Accelerating AutoDock4 with GPUs and Gradient-Based Local Search.

AutoDock4 is a widely used program for docking small molecules to macromolecular targets. It describes ligand-receptor interactions using a physics-inspired scoring function that has been proven useful in a variety of drug discovery projects. However, compared to more modern and recent software, AutoDock4 has longer execution times, limiting its applicability to large scale dockings. To address this problem, we describe an OpenCL implementation of AutoDock4, called AutoDock-GPU, that leverages the highly parallel architecture of GPU hardware to reduce docking runtime by up to 350-fold with respect to a single-threaded process. Moreover, we introduce the gradient-based local search method ADADELTA, as well as an improved version of the Solis-Wets random optimizer from AutoDock4. These efficient local search algorithms significantly reduce the number of calls to the scoring function that are needed to produce good results. The improvements reported here, both in terms of docking throughput and search efficiency, facilitate the use of the AutoDock4 scoring function in large scale virtual screening.

C-terminal residues of activated protein C light chain contribute to its anticoagulant and cytoprotective activities.

BACKGROUND: Activated protein C (APC) is an important homeostatic blood coagulation protease that conveys anticoagulant and cytoprotective activities. Proteolytic inactivation of factors Va and VIIIa facilitated by cofactor protein S is responsible for APC's anticoagulant effects, whereas cytoprotective effects of APC involve primarily the endothelial protein C receptor (EPCR), protease activated receptor (PAR)1 and PAR3. OBJECTIVE: To date, several binding exosites in the protease domain of APC have been identified that contribute to APC's interaction with its substrates but potential contributions of the C-terminus of the light chain have not been studied in detail. METHODS: Site-directed Ala-scanning mutagenesis of six positively charged residues within G142-L155 was used to characterize their contributions to APC's anticoagulant and cytoprotective activities. RESULTS AND CONCLUSIONS: K151 was involved in protein S dependent-anticoagulant activity of APC with some contribution of K150. 3D structural analysis supported that these two residues were exposed in an extended protein S binding site on one face of APC. Both K150 and K151 were important for PAR1 and PAR3 cleavage by APC, suggesting that this region may also mediate interactions with PARs. Accordingly, APC's cytoprotective activity as determined by endothelial barrier protection was impaired by Ala substitutions of these residues. Thus, both K150 and K151 are involved in APC's anticoagulant and cytoprotective activities. The differential contribution of K150 relative to K151 for protein S-dependent anticoagulant activity and PAR cleavage highlights that binding exosites for protein S binding and for PAR cleavage in the C-terminal region of APC's light chain overlap.

AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility.

We describe the testing and release of AutoDock4 and the accompanying graphical user interface AutoDockTools. AutoDock4 incorporates limited flexibility in the receptor. Several tests are reported here, including a redocking experiment with 188 diverse ligand-protein complexes and a cross-docking experiment using flexible sidechains in 87 HIV protease complexes. We also report its utility in analysis of covalently bound ligands, using both a grid-based docking method and a modification of the flexible sidechain technique.

Docking Flexible Cyclic Peptides with AutoDock CrankPep.

While a new therapeutic cyclic peptide is approved nearly every year, docking large macrocycles has remained challenging. Here, we present a new version of our peptide docking software AutoDock CrankPep (ADCP), extended to dock peptides cyclized through their backbone and/or side chain disulfide bonds. We show that within the top 10 solutions, ADCP identifies the proper interactions for 71% of a data set of 38 complexes, thus making it a useful tool for rational peptide-based drug design.

FLIPDock: docking flexible ligands into flexible receptors.

Conformational changes of biological macromolecules when binding with ligands have long been observed and remain a challenge for automated docking methods. Here we present a novel protein-ligand docking software called FLIPDock (Flexible LIgand-Protein Docking) allowing the automated docking of flexible ligand molecules into active sites of flexible receptor molecules. In FLIPDock, conformational spaces of molecules are encoded using a data structure that we have developed recently called the Flexibility Tree (FT). While the FT can represent fully flexible ligands, it was initially designed as a hierarchical and multiresolution data structure for the selective encoding of conformational subspaces of large biological macromolecules. These conformational subspaces can be built to span a range of conformations important for the biological activity of a protein. A variety of motions can be combined, ranging from domains moving as rigid bodies or backbone atoms undergoing normal mode-based deformations, to side chains assuming rotameric conformations. In addition, these conformational subspaces are parameterized by a small number of variables which can be searched during the docking process, thus effectively modeling the conformational changes in a flexible receptor. FLIPDock searches the variables using genetic algorithm-based search techniques and evaluates putative docking complexes with a scoring function based on the AutoDock3.05 force-field. In this paper, we describe the concepts behind FLIPDock and the overall architecture of the program. We demonstrate FLIPDock's ability to solve docking problems in which the assumption of a rigid receptor previously prevented the successful docking of known ligands. In particular, we repeat an earlier cross docking experiment and demonstrate an increased success rate of 93.5%, compared to original 72% success rate achieved by AutoDock over the 400 cross-docking calculations. We also demonstrate FLIPDock's ability to handle conformational changes involving backbone motion by docking balanol to an adenosine-binding pocket of protein kinase A.

A component-based software environment for visualizing large macromolecular assemblies.

The interactive visualization of large biological assemblies poses a number of challenging problems, including the development of multiresolution representations and new interaction methods for navigating and analyzing these complex systems. An additional challenge is the development of flexible software environments that will facilitate the integration and interoperation of computational models and techniques from a wide variety of scientific disciplines. In this paper, we present a component-based software development strategy centered on the high-level, object-oriented, interpretive programming language: Python. We present several software components, discuss their integration, and describe some of their features that are relevant to the visualization of large molecular assemblies. Several examples are given to illustrate the interoperation of these software components and the integration of structural data from a variety of experimental sources. These examples illustrate how combining visual programming with component-based software development facilitates the rapid prototyping of novel visualization tools.

Activated protein C light chain provides an extended binding surface for its anticoagulant cofactor, protein S.

Protein S anticoagulant cofactor sensitivity and PAR1 cleavage activity were assayed for 9 recombinant APC mutants.Residues L38, K43, I73, F95, and W115 on one face of the APC light chain define an extended surface containing the protein S binding site.

Computational protein-ligand docking and virtual drug screening with the AutoDock suite.

Computational docking can be used to predict bound conformations and free energies of binding for small-molecule ligands to macromolecular targets. Docking is widely used for the study of biomolecular interactions and mechanisms, and it is applied to structure-based drug design. The methods are fast enough to allow virtual screening of ligand libraries containing tens of thousands of compounds. This protocol covers the docking and virtual screening methods provided by the AutoDock suite of programs, including a basic docking of a drug molecule with an anticancer target, a virtual screen of this target with a small ligand library, docking with selective receptor flexibility, active site prediction and docking with explicit hydration. The entire protocol will require ∼5 h.

Evolutionary analysis of HIV-1 protease inhibitors: Methods for design of inhibitors that evade resistance.

Drug-resistant strains are rapidly selected during AIDS therapy because of the high rate of mutation in HIV. In this report, we present an evolutionary simulation method for analysis of viral mutation and its use for optimization of HIV-1 protease drugs to improve their robustness in the face of resistance mutation. We first present an analysis of the range of resistant mutants that produce viable viruses by using a volume-based viral fitness model. Then, we analyze how this range of mutant proteases allows development of resistance to an optimal inhibitor previously designed by computational coevolution techniques. Finally, we evaluate the resistance patterns of commercially available drugs, and we discuss how resistance might be overcome by optimizing the size of specific side-chains of these inhibitors.

Model of the alphaLbeta2 integrin I-domain/ICAM-1 DI interface suggests that subtle changes in loop orientation determine ligand specificity.

The interaction of the alphaLbeta2 integrin with its cellular ligand the intercellular adhesion molecule-1 (ICAM-1) is critical for the tight binding interaction between most leukocytes and the vascular endothelium before transendothelial migration to the sites of inflammation. In this article we have modeled the alphaL subunit I-domain in its active form, which was computationally docked with the D1 domain of the ICAM-1 to probe potential protein-protein interactions. The experimentally observed key interaction between the carboxylate of Glu 34 in the ICAM-1 D1 domain and the metal ion-dependent adhesion site (MIDAS) in the open alphaL I-domain was consistently reproduced by our calculations. The calculations reveal the nature of the alphaLbeta2/ICAM-1 interaction and suggest an explanation for the increased ligand-binding affinity in the "open" versus the "closed" conformation of the alphaL I-domain. A mechanism for substrate selectivity among alphaL, alphaM, and alpha2 I-domains is suggested whereby the orientation of the loops within the I-domain is critical in mediating the interaction of the Glu 34 carboxylate of ICAM-1 D1 with the MIDAS.

3D molecular models of whole HIV-1 virions generated with cellPACK.

As knowledge of individual biological processes grows, it becomes increasingly useful to frame new findings within their larger biological contexts in order to generate new systems-scale hypotheses. This report highlights two major iterations of a whole virus model of HIV-1, generated with the cellPACK software. cellPACK integrates structural and systems biology data with packing algorithms to assemble comprehensive 3D models of cell-scale structures in molecular detail. This report describes the biological data, modeling parameters and cellPACK methods used to specify and construct editable models for HIV-1. Anticipating that cellPACK interfaces under development will enable researchers from diverse backgrounds to critique and improve the biological models, we discuss how cellPACK can be used as a framework to unify different types of data across all scales of biology.