Rosetta FlexPepDock Research Articles

Design of target specific peptide inhibitors using generative deep learning and molecular dynamics simulations

We introduce a computational approach for the design of target-specific peptides. Our method integrates a Gated Recurrent Unit-based Variational Autoencoder with Rosetta FlexPepDock for peptide sequence generation and binding affinity assessment. Subsequently, molecular dynamics simulations are employed to narrow down the selection of peptides for experimental assays. We apply this computational strategy to design peptide inhibitors that specifically target β-catenin and NF-κB essential modulator. Among the twelve β-catenin inhibitors, six exhibit improved binding affinity compared to the parent peptide. Notably, the best C-terminal peptide binds β-catenin with an IC50 of 0.010 ± 0.06 μM, which is 15-fold better than the parent peptide. For NF-κB essential modulator, two of the four tested peptides display substantially enhanced binding compared to the parent peptide. Collectively, this study underscores the successful integration of deep learning and structure-based modeling and simulation for target specific peptide design.

Use of phosphotyrosine-containing peptides to target SH2 domains: Antagonist peptides of the Crk/CrkL-p130Cas axis.

Protein-protein interactions between SH2 domains and segments of proteins that include a post-translationally phosphorylated tyrosine residue (pY) underpin numerous signal transduction cascades that allow cells to respond to their environment. Dysregulation of the writing, erasing, and reading of these posttranslational modifications is a hallmark of human disease, notably cancer. Elucidating the precise role of the SH2 domain-containing adaptor proteins Crk and CrkL in tumor cell migration and invasion is challenging because there are no specific and potent antagonists available. Crk and CrkL SH2s interact with a region of the docking protein p130Cas containing 15 potential pY-containing tetrapeptide motifs. This chapter summarizes recent efforts toward peptide antagonists for this Crk/CrkL-p130Cas interaction. We describe our protocol for recombinant expression and purification of Crk and CrkL SH2s for functional assays and our procedure to determine the consensus binding motif from the p130Cas sequence. To develop a more potent antagonist, we employ methods often associated with structure-based drug design. Computational docking using Rosetta FlexPepDock, which accounts for peptides having a greater number of conformational degrees of freedom than small organic molecules that typically constitute libraries, provides quantitative docking metrics to prioritize candidate peptides for experimental testing. A battery of biophysical assays, including fluorescence polarization, differential scanning fluorimetry and saturation transfer difference nuclear magnetic resonance spectroscopy, were employed to assess the candidates. In parallel, GST pulldown competition assays characterized protein-protein binding in vitro. Taken together, our methodology yields peptide antagonists of the Crk/CrkL-p130Cas axis that will be used to validate targets, assess druggability, foster in vitro assay development, and potentially serve as lead compounds for therapeutic intervention.

Elucidation of Short Linear Motif-Based Interactions of the FERM Domains of Ezrin, Radixin, Moesin, and Merlin.

The ERM (ezrin, radixin, and moesin) family of proteins and the related protein merlin participate in scaffolding and signaling events at the cell cortex. The proteins share an N-terminal FERM [band four-point-one (4.1) ERM] domain composed of three subdomains (F1, F2, and F3) with binding sites for short linear peptide motifs. By screening the FERM domains of the ERMs and merlin against a phage library that displays peptides representing the intrinsically disordered regions of the human proteome, we identified a large number of novel ligands. We determined the affinities for the ERM and merlin FERM domains interacting with 18 peptides and validated interactions with full-length proteins through pull-down experiments. The majority of the peptides contained an apparent Yx[FILV] motif; others show alternative motifs. We defined distinct binding sites for two types of similar but distinct binding motifs (YxV and FYDF) using a combination of Rosetta FlexPepDock computational peptide docking protocols and mutational analysis. We provide a detailed molecular understanding of how the two types of peptides with distinct motifs bind to different sites on the moesin FERM phosphotyrosine binding-like subdomain and uncover interdependencies between the different types of ligands. The study expands the motif-based interactomes of the ERMs and merlin and suggests that the FERM domain acts as a switchable interaction hub.

Rosetta FlexPepDock to predict peptide-MHC binding: An approach for non-canonical amino acids.

Computation methods that predict the binding of peptides to MHC-I are important tools for screening and identifying immunogenic antigens and have the potential to accelerate vaccine and drug development. However, most available tools are sequence-based and optimized only for peptides containing the twenty canonical amino acids. This omits a large number of peptides containing non-canonical amino acids (NCAA), or residues that undergo varied post-translational modifications such as glycosylation or phosphorylation. These modifications fundamentally alter peptide immunogenicity. Similarly, existing structure-based methods are biased towards canonical peptide backbone structures, which may or may not be preserved when NCAAs are present. Rosetta FlexPepDock ab-initio is a structure-based computational protocol able to evaluate peptide-receptor interaction where no prior information of the peptide backbone is known. We benchmarked FlexPepDock ab-initio for docking canonical peptides to MHC-I, and illustrate for the first time the method's ability to accurately model MHC-I bound epitopes containing NCAAs. FlexPepDock ab-initio protocol was able to recapitulate near-native structures (≤1.5Å) in the top lowest-energy models for 20 out of 25 cases in our initial benchmark. Using known experimental binding affinities of twenty peptides derived from an influenza-derived peptide, we showed that FlexPepDock protocol is able to predict relative binding affinity as Rosetta energies correlate well with experimental values (r = 0.59, p = 0.006). ROC analysis revealed 80% true positive and a 40% false positive rate, with a prediction power of 93%. Finally, we demonstrate the protocol's ability to accurately recapitulate HLA-A*02:01 bound phosphopeptide backbone structures and relative binding affinity changes, the theoretical structure of the lymphocytic choriomeningitis derived glycosylated peptide GP392 bound to MHC-I H-2Db, and isolevuglandin-adducted peptides. The ability to use non-canonical amino acids in the Rosetta FlexPepDock protocol may provide useful insight into critical amino acid positions where the post-translational modification modulates immunologic responses.

InterPepScore: a deep learning score for improving the FlexPepDock refinement protocol.

MotivationInteractions between peptide fragments and protein receptors are vital to cell function yet difficult to experimentally determine in structural details of. As such, many computational methods have been developed to aid in peptide–protein docking or structure prediction. One such method is Rosetta FlexPepDock which consistently refines coarse peptide–protein models into sub-Ångström precision using Monte-Carlo simulations and statistical potentials. Deep learning has recently seen increased use in protein structure prediction, with graph neural networks used for protein model quality assessment.ResultsHere, we introduce a graph neural network, InterPepScore, as an additional scoring term to complement and improve the Rosetta FlexPepDock refinement protocol. InterPepScore is trained on simulation trajectories from FlexPepDock refinement starting from thousands of peptide–protein complexes generated by a wide variety of docking schemes. The addition of InterPepScore into the refinement protocol consistently improves the quality of models created, and on an independent benchmark on 109 peptide–protein complexes its inclusion results in an increase in the number of complexes for which the top-scoring model had a DockQ-score of 0.49 (Medium quality) or better from 14.8% to 26.1%.Availability and implementationInterPepScore is available online at http://wallnerlab.org/InterPepScore.Supplementary information Supplementary data are available at Bioinformatics online.

Combining Three-Dimensional Modeling with Artificial Intelligence to Increase Specificity and Precision in Peptide-MHC Binding Predictions.

The reliable prediction of the affinity of candidate peptides for the MHC is important for predicting their potential antigenicity and thus influences medical applications, such as decisions on their inclusion in T cell-based vaccines. In this study, we present a rapid, predictive computational approach that combines a popular, sequence-based artificial neural network method, NetMHCpan 4.0, with three-dimensional structural modeling. We find that the ensembles of bound peptide conformations generated by the programs MODELLER and Rosetta FlexPepDock are less variable in geometry for strong binders than for low-affinity peptides. In tests on 1271 peptide sequences for which the experimental dissociation constants of binding to the well-characterized murine MHC allele H-2Db are known, by applying thresholds for geometric fluctuations the structure-based approach in a standalone manner drastically improves the statistical specificity, reducing the number of false positives. Furthermore, filtering candidates generated with NetMHCpan 4.0 with the structure-based predictor led to an increase in the positive predictive value (PPV) of the peptides correctly predicted to bind very strongly (i.e., K d < 100 nM) from 40 to 52% (p = 0.027). The combined method also significantly improved the PPV when tested on five human alleles, including some with limited data for training. Overall, an average increase of 10% in the PPV was found over the standalone sequence-based method. The combined method should be useful in the rapid design of effective T cell-based vaccines.

Docking of peptides to GPCRs using a combination of CABS-dock with FlexPepDock refinement.

The structural description of peptide ligands bound to G protein-coupled receptors (GPCRs) is important for the discovery of new drugs and deeper understanding of the molecular mechanisms of life. Here we describe a three-stage protocol for the molecular docking of peptides to GPCRs using a set of different programs: (1) CABS-dock for docking fully flexible peptides; (2) PD2 method for the reconstruction of atomistic structures from C-alpha traces provided by CABS-dock and (3) Rosetta FlexPepDock for the refinement of protein–peptide complex structures and model scoring. We evaluated the proposed protocol on the set of seven different GPCR–peptide complexes (including one containing a cyclic peptide), for which crystallographic structures are available. We show that CABS-dock produces high resolution models in the sets of top-scored models. These sets of models, after reconstruction to all-atom representation, can be further improved by Rosetta high-resolution refinement and/or minimization, leading in most of the cases to sub-Angstrom accuracy in terms of interface root-mean-square-deviation measure.

Modeling beta-sheet peptide-protein interactions: Rosetta FlexPepDock in CAPRI rounds 38-45.

Peptide-protein docking is challenging due to the considerable conformational freedom of the peptide. CAPRI rounds 38-45 included two peptide-protein interactions, both characterized by a peptide forming an additional beta strand of a beta sheet in the receptor. Using the Rosetta FlexPepDock peptide docking protocol we generated top-performing, high-accuracy models for targets 134 and 135, involving an interaction between a peptide derived from L-MAG with DLC8. In addition, we were able to generate the only medium-accuracy models for a particularly challenging target, T121. In contrast to the classical peptide-mediated interaction, in which receptor side chains contact both peptide backbone and side chains, beta-sheet complementation involves a major contribution to binding by hydrogen bonds between main chain atoms. To establish how binding affinity and specificity are established in this special class of peptide-protein interactions, we extracted PeptiDBeta, a benchmark of solved structures of different protein domains that are bound by peptides via beta-sheet complementation, and tested our protocol for global peptide-docking PIPER-FlexPepDock on this dataset. We find that the beta-strand part of the peptide is sufficient to generate approximate and even high resolution models of many interactions, but inclusion of adjacent motif residues often provides additional information necessary to achieve high resolution model quality.

Protocols for All-Atom Reconstruction and High-Resolution Refinement of Protein-Peptide Complex Structures.

Structural characterizations of protein-peptide complexes may require further improvements. These may include reconstruction of missing atoms and/or structure optimization leading to higher accuracy models. In this work, we describe a workflow that generates accurate structural models of peptide-protein complexes starting from protein-peptide models in C-alpha representation generated using CABS-dock molecular docking. First, protein-peptide models are reconstructed from their C-alpha traces to all-atom representation using MODELLER. Next, they are refined using Rosetta FlexPepDock. The described workflow allows for reliable all-atom reconstruction of CABS-dock models and their further improvement to high-resolution models.

Systematically benchmarking peptide-MHC binding predictors: From synthetic to naturally processed epitopes.

A number of machine learning-based predictors have been developed for identifying immunogenic T-cell epitopes based on major histocompatibility complex (MHC) class I and II binding affinities. Rationally selecting the most appropriate tool has been complicated by the evolving training data and machine learning methods. Despite the recent advances made in generating high-quality MHC-eluted, naturally processed ligandome, the reliability of new predictors on these epitopes has yet to be evaluated. This study reports the latest benchmarking on an extensive set of MHC-binding predictors by using newly available, untested data of both synthetic and naturally processed epitopes. 32 human leukocyte antigen (HLA) class I and 24 HLA class II alleles are included in the blind test set. Artificial neural network (ANN)-based approaches demonstrated better performance than regression-based machine learning and structural modeling. Among the 18 predictors benchmarked, ANN-based mhcflurry and nn_align perform the best for MHC class I 9-mer and class II 15-mer predictions, respectively, on binding/non-binding classification (Area Under Curves = 0.911). NetMHCpan4 also demonstrated comparable predictive power. Our customization of mhcflurry to a pan-HLA predictor has achieved similar accuracy to NetMHCpan. The overall accuracy of these methods are comparable between 9-mer and 10-mer testing data. However, the top methods deliver low correlations between the predicted versus the experimental affinities for strong MHC binders. When used on naturally processed MHC-ligands, tools that have been trained on elution data (NetMHCpan4 and MixMHCpred) shows better accuracy than pure binding affinity predictor. The variability of false prediction rate is considerable among HLA types and datasets. Finally, structure-based predictor of Rosetta FlexPepDock is less optimal compared to the machine learning approaches. With our benchmarking of MHC-binding and MHC-elution predictors using a comprehensive metrics, a unbiased view for establishing best practice of T-cell epitope predictions is presented, facilitating future development of methods in immunogenomics.

High-resolution global peptide-protein docking using fragments-based PIPER-FlexPepDock.

Peptide-protein interactions contribute a significant fraction of the protein-protein interactome. Accurate modeling of these interactions is challenging due to the vast conformational space associated with interactions of highly flexible peptides with large receptor surfaces. To address this challenge we developed a fragment based high-resolution peptide-protein docking protocol. By streamlining the Rosetta fragment picker for accurate peptide fragment ensemble generation, the PIPER docking algorithm for exhaustive fragment-receptor rigid-body docking and Rosetta FlexPepDock for flexible full-atom refinement of PIPER docked models, we successfully addressed the challenge of accurate and efficient global peptide-protein docking at high-resolution with remarkable accuracy, as validated on a small but representative set of peptide-protein complex structures well resolved by X-ray crystallography. Our approach opens up the way to high-resolution modeling of many more peptide-protein interactions and to the detailed study of peptide-protein association in general. PIPER-FlexPepDock is freely available to the academic community as a server at http://piperfpd.furmanlab.cs.huji.ac.il.

FlexPepDock lessons from CAPRI peptide-protein rounds and suggested new criteria for assessment of model quality and utility.

ABSTRACTCAPRI rounds 28 and 29 included, for the first time, peptide‐receptor targets of three different systems, reflecting increased appreciation of the importance of peptide‐protein interactions. The CAPRI rounds allowed us to objectively assess the performance of Rosetta FlexPepDock, one of the first protocols to explicitly include peptide flexibility in docking, accounting for peptide conformational changes upon binding. We discuss here successes and challenges in modeling these targets: we obtain top‐performing, high‐resolution models of the peptide motif for cases with known binding sites but there is a need for better modeling of flanking regions, as well as better selection criteria, in particular for unknown binding sites. These rounds have also provided us the opportunity to reassess the success criteria, to better reflect the quality of a peptide‐protein complex model. Using all models submitted to CAPRI, we analyze the correlation between current classification criteria and the ability to retrieve critical interface features, such as hydrogen bonds and hotspots. We find that loosening the backbone (and ligand) RMSD threshold, together with a restriction on the side chain RMSD measure, allows us to improve the selection of high‐accuracy models. We also suggest a new measure to assess interface hydrogen bond recovery, which is not assessed by the current CAPRI criteria. Finally, we find that surprisingly much can be learned from rather inaccurate models about binding hotspots, suggesting that the current status of peptide–protein docking methods, as reflected by the submitted CAPRI models, can already have a significant impact on our understanding of protein interactions. Proteins 2017; 85:445–462. © 2016 The Authors Proteins: Structure, Function, and Bioinformatics Published by Wiley Periodicals, Inc.

Modeling Peptide-Protein Structure and Binding Using Monte Carlo Sampling Approaches: Rosetta FlexPepDock and FlexPepBind.

Many signaling and regulatory processes involve peptide-mediated protein interactions, i.e., the binding of a short stretch in one protein to a domain in its partner. Computational tools that generate accurate models of peptide-receptor structures and binding improve characterization and manipulation of known interactions, help to discover yet unknown peptide-protein interactions and networks, and bring into reach the design of peptide-based drugs for targeting specific systems of medical interest.Here, we present a concise overview of the Rosetta FlexPepDock protocol and its derivatives that we have developed for the structure-based characterization of peptide-protein binding. Rosetta FlexPepDock was built to generate precise models of protein-peptide complex structures, by effectively addressing the challenge of the considerable conformational flexibility of the peptide. Rosetta FlexPepBind is an extension of this protocol that allows characterizing peptide-binding affinities and specificities of various biological systems, based on the structural models generated by Rosetta FlexPepDock. We provide detailed descriptions and guidelines for the usage of these protocols, and on a specific example, we highlight the variety of different challenges that can be met and the questions that can be answered with Rosetta FlexPepDock.

Subangstrom accuracy in pHLA-I modeling by Rosetta FlexPepDock refinement protocol.

Flexible peptides binding to human leukocyte antigen (HLA) play a key role in mediating human immune responses and are also involved in idiosyncratic adverse drug reactions according to recent research. However, the structural determinations of pHLA complexes remain challenging under the present conditions. In this paper, the performance of a new peptide docking method, namely FlexPepDock, was systematically investigated by a benchmark of 30 crystallized structures of peptide-HLA class I complexes. The docking results showed that the near-native pHLA-I models with peptide bb-RMSD less than 2 Å were ranked in the top 1 model for 100% (70/70) docking cases, and the subangstrom models with peptide bb-RMSD less than 1 Å were ranked in the top 5 lowest-energy models for 65.7% (46/70) docking cases. Furthermore, 10 out of 70 docking cases ranked the subangstrom all-atom models in the top 5 lowest-energy models. The results showed that the FlexPepDock can generate high-quality models of pHLA-I complexes and can be widely applied to pHLA-I modeling and mechanism research of peptide-mediated immune responses.

Biomimetic Design of Affinity Peptide Ligand for Capsomere of Virus-Like Particle

Virus-like particle (VLP) of murine polyomavirus (MPV) is a T = 7d icosahedral capsid that self-assembles from 72 capsomeres (Caps), each of which is a pentamer of major coat protein VP1. VLP has great potential in vaccinology, gene therapy, drug delivery, and materials science. However, its application is hindered by high cost downstream processes, leading to an urgent demand of a highly efficient affinity ligand for the separation and purification of Cap by affinity chromatography. Herein a biomimetic design strategy of an affinity peptide ligand of Cap has been developed on the basis of the binding structure of the C-terminus of minor coat protein (VP2-C) on the inner surface of Cap. The molecular interactions between VP2-C and Cap were first examined using all-atom molecular dynamics (MD) simulations coupled with the molecular mechanics/Poisson-Boltzmann surface area (MM/PBSA) method, where V283, P285, D286, W287, L289, and Y296 of VP2-C were identified as the hot spots. An affinity peptide library (DWXLXLXY, X denotes arbitrary amino acids except cysteine) was then constructed for virtual screening sequently by docking with AUTODOCK VINA, binding structure comparison, and final docking with ROSETTA FlexPepDock. Ten peptide candidates were selected and further confirmed by MD simulations and MM/PBSA, where DWDLRLLY was found to have the highest affinity to Cap. In DWDLRLLY, six residues are favorable for the binding, including W2, L4, L6 and Y8 inheriting from VP2-C, and R5 and L7 selected in the virtual screening. This confirms the high efficiency and accuracy of the biomimetic design strategy. DWDLRLLY was then experimentally validated by a one-step purification of Cap from crude cell lysate using affinity chromatography with the octapeptide immobilized on Sepharose gel. The purified Caps were observed to self-assemble into VLP with consistent structure of authentic MPV.

Improved Docking of Polypeptides with Glide

Predicting the binding mode of flexible polypeptides to proteins is an important task that falls outside the domain of applicability of most small molecule and protein-protein docking tools. Here, we test the small molecule flexible ligand docking program Glide on a set of 19 non-α-helical peptides and systematically improve pose prediction accuracy by enhancing Glide sampling for flexible polypeptides. In addition, scoring of the poses was improved by post-processing with physics-based implicit solvent MM-GBSA calculations. Using the best RMSD among the top 10 scoring poses as a metric, the success rate (RMSD ≤ 2.0 Å for the interface backbone atoms) increased from 21% with default Glide SP settings to 58% with the enhanced peptide sampling and scoring protocol in the case of redocking to the native protein structure. This approaches the accuracy of the recently developed Rosetta FlexPepDock method (63% success for these 19 peptides) while being over 100 times faster. Cross-docking was performed for a subset of cases where an unbound receptor structure was available, and in that case, 40% of peptides were docked successfully. We analyze the results and find that the optimized polypeptide protocol is most accurate for extended peptides of limited size and number of formal charges, defining a domain of applicability for this approach.

Rosetta FlexPepDock web server—high resolution modeling of peptide–protein interactions

Peptide–protein interactions are among the most prevalent and important interactions in the cell, but a large fraction of those interactions lack detailed structural characterization. The Rosetta FlexPepDock web server (http://flexpepdock.furmanlab.cs.huji.ac.il/) provides an interface to a high-resolution peptide docking (refinement) protocol for the modeling of peptide–protein complexes, implemented within the Rosetta framework. Given a protein receptor structure and an approximate, possibly inaccurate model of the peptide within the receptor binding site, the FlexPepDock server refines the peptide to high resolution, allowing full flexibility to the peptide backbone and to all side chains. This protocol was extensively tested and benchmarked on a wide array of non-redundant peptide–protein complexes, and was proven effective when applied to peptide starting conformations within 5.5 Å backbone root mean square deviation from the native conformation. FlexPepDock has been applied to several systems that are mediated and regulated by peptide–protein interactions. This easy to use and general web server interface allows non-expert users to accurately model their specific peptide–protein interaction of interest.

Rosetta FlexPepDock ab-initio: simultaneous folding, docking and refinement of peptides onto their receptors.

Flexible peptides that fold upon binding to another protein molecule mediate a large number of regulatory interactions in the living cell and may provide highly specific recognition modules. We present Rosetta FlexPepDock ab-initio, a protocol for simultaneous docking and de-novo folding of peptides, starting from an approximate specification of the peptide binding site. Using the Rosetta fragments library and a coarse-grained structural representation of the peptide and the receptor, FlexPepDock ab-initio samples efficiently and simultaneously the space of possible peptide backbone conformations and rigid-body orientations over the receptor surface of a given binding site. The subsequent all-atom refinement of the coarse-grained models includes full side-chain modeling of both the receptor and the peptide, resulting in high-resolution models in which key side-chain interactions are recapitulated. The protocol was applied to a benchmark in which peptides were modeled over receptors in either their bound backbone conformations or in their free, unbound form. Near-native peptide conformations were identified in 18/26 of the bound cases and 7/14 of the unbound cases. The protocol performs well on peptides from various classes of secondary structures, including coiled peptides with unusual turns and kinks. The results presented here significantly extend the scope of state-of-the-art methods for high-resolution peptide modeling, which can now be applied to a wide variety of peptide-protein interactions where no prior information about the peptide backbone conformation is available, enabling detailed structure-based studies and manipulation of those interactions.

Autophosphorylation Activates Dictyostelium Myosin II Heavy Chain Kinase A by Providing a Ligand for an Allosteric Binding Site in the α-Kinase Domain

Dictyostelium discoideum myosin II heavy chain kinase A (MHCK A), a member of the atypical α-kinase family, phosphorylates sites in the myosin II tail that block filament assembly. Here we show that the catalytic activity of A-CAT, the α-kinase domain of MHCK A (residues 552-841), is severely inhibited by the removal of a disordered C-terminal tail sequence (C-tail; residues 806-841). The key residue in the C-tail was identified as Thr(825), which was found to be constitutively autophosphorylated. Dephosphorylation of Thr(825) using shrimp alkaline phosphatase decreased A-CAT activity. The activity of a truncated A-CAT lacking Thr(825) could be rescued by P(i), phosphothreonine, and a phosphorylated peptide, but not by threonine, glutamic acid, aspartic acid, or an unphosphorylated peptide. These results focused attention on a P(i)-binding pocket located in the C-terminal lobe of A-CAT. Mutational analysis demonstrated that the P(i)-pocket was essential for A-CAT activity. Based on these results, it is proposed that autophosphorylation of Thr(825) activates ACAT by providing a covalently tethered ligand for the P(i)-pocket. Ab initio modeling studies using the Rosetta FloppyTail and FlexPepDock protocols showed that it is feasible for the phosphorylated Thr(825) to dock intramolecularly into the P(i)-pocket. Allosteric activation is predicted to involve a conformational change in Arg(734), which bridges the bound P(i) to Asp(762) in a key active site loop. Sequence alignments indicate that a comparable regulatory mechanism is likely to be conserved in Dictyostelium MHCK B-D and metazoan eukaryotic elongation factor-2 kinases.

Sub-angstrom modeling of complexes between flexible peptides and globular proteins

A wide range of regulatory processes in the cell are mediated by flexible peptides that fold upon binding to globular proteins. Computational efforts to model these interactions are hindered by the large number of rotatable bonds in flexible peptides relative to typical ligand molecules, and the fact that different peptides assume different backbone conformations within the same binding site. In this study, we present Rosetta FlexPepDock, a novel tool for refining coarse peptide-protein models that allows significant changes in both peptide backbone and side chains. We obtain high resolution models, often of sub-angstrom backbone quality, over an extensive and general benchmark that is based on a large nonredundant dataset of 89 peptide-protein interactions. Importantly, side chains of known binding motifs are modeled particularly well, typically with atomic accuracy. In addition, our protocol has improved modeling quality for the important application of cross docking to PDZ domains. We anticipate that the ability to create high resolution models for a wide range of peptide-protein complexes will have significant impact on structure-based functional characterization, controlled manipulation of peptide interactions, and on peptide-based drug design.