DeepRank-GNN: a graph neural network framework to learn patterns in protein-protein interfaces.

MOTIVATION: Gaining structural insights into the protein-protein interactome is essential to understand biological phenomena and extract knowledge for rational drug design or protein engineering. We have previously developed DeepRank, a deep-learning framework to facilitate pattern learning from protein-protein interfaces using convolutional neural network (CNN) approaches. However, CNN is not rotation invariant and data augmentation is required to desensitize the network to the input data orientation which dramatically impairs the computation performance. Representing protein-protein complexes as atomic- or residue-scale rotation invariant graphs instead enables using graph neural networks (GNN) approaches, bypassing those limitations. RESULTS: We have developed DeepRank-GNN, a framework that converts protein-protein interfaces from PDB 3D coordinates files into graphs that are further provided to a pre-defined or user-defined GNN architecture to learn problem-specific interaction patterns. DeepRank-GNN is designed to be highly modularizable, easily customized and is wrapped into a user-friendly python3 package. Here, we showcase DeepRank-GNN's performance on two applications using a dedicated graph interaction neural network: (i) the scoring of docking poses and (ii) the discriminating of biological and crystal interfaces. In addition to the highly competitive performance obtained in those tasks as compared to state-of-the-art methods, we show a significant improvement in speed and storage requirement using DeepRank-GNN as compared to DeepRank. AVAILABILITY AND IMPLEMENTATION: DeepRank-GNN is freely available from https://github.com/DeepRank/DeepRank-GNN. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Discriminating physiological from non-physiological interfaces in structures of protein complexes: A community-wide study.

Reliably scoring and ranking candidate models of protein complexes and assigning their oligomeric state from the structure of the crystal lattice represent outstanding challenges. A community-wide effort was launched to tackle these challenges. The latest resources on protein complexes and interfaces were exploited to derive a benchmark dataset consisting of 1677 homodimer protein crystal structures, including a balanced mix of physiological and non-physiological complexes. The non-physiological complexes in the benchmark were selected to bury a similar or larger interface area than their physiological counterparts, making it more difficult for scoring functions to differentiate between them. Next, 252 functions for scoring protein-protein interfaces previously developed by 13 groups were collected and evaluated for their ability to discriminate between physiological and non-physiological complexes. A simple consensus score generated using the best performing score of each of the 13 groups, and a cross-validated Random Forest (RF) classifier were created. Both approaches showed excellent performance, with an area under the Receiver Operating Characteristic (ROC) curve of 0.93 and 0.94, respectively, outperforming individual scores developed by different groups. Additionally, AlphaFold2 engines recalled the physiological dimers with significantly higher accuracy than the non-physiological set, lending support to the reliability of our benchmark dataset annotations. Optimizing the combined power of interface scoring functions and evaluating it on challenging benchmark datasets appears to be a promising strategy.

Prediction of protein assemblies, the next frontier: The CASP14-CAPRI experiment.

We present the results for CAPRI Round 50, the fourth joint CASP-CAPRI protein assembly prediction challenge. The Round comprised a total of twelve targets, including six dimers, three trimers, and three higher-order oligomers. Four of these were easy targets, for which good structural templates were available either for the full assembly, or for the main interfaces (of the higher-order oligomers). Eight were difficult targets for which only distantly related templates were found for the individual subunits. Twenty-five CAPRI groups including eight automatic servers submitted ~1250 models per target. Twenty groups including six servers participated in the CAPRI scoring challenge submitted ~190 models per target. The accuracy of the predicted models was evaluated using the classical CAPRI criteria. The prediction performance was measured by a weighted scoring scheme that takes into account the number of models of acceptable quality or higher submitted by each group as part of their five top-ranking models. Compared to the previous CASP-CAPRI challenge, top performing groups submitted such models for a larger fraction (70-75%) of the targets in this Round, but fewer of these models were of high accuracy. Scorer groups achieved stronger performance with more groups submitting correct models for 70-80% of the targets or achieving high accuracy predictions. Servers performed less well in general, except for the MDOCKPP and LZERD servers, who performed on par with human groups. In addition to these results, major advances in methodology are discussed, providing an informative overview of where the prediction of protein assemblies currently stands.

Shape-Restrained Modeling of Protein-Small-Molecule Complexes with High Ambiguity Driven DOCKing.

Small-molecule docking remains one of the most valuable computational techniques for the structure prediction of protein-small-molecule complexes. It allows us to study the interactions between compounds and the protein receptors they target at atomic detail in a timely and efficient manner. Here, we present a new protocol in HADDOCK (High Ambiguity Driven DOCKing), our integrative modeling platform, which incorporates homology information for both receptor and compounds. It makes use of HADDOCK's unique ability to integrate information in the simulation to drive it toward conformations, which agree with the provided data. The focal point is the use of shape restraints derived from homologous compounds bound to the target receptors. We have developed two protocols: in the first, the shape is composed of dummy atom beads based on the position of the heavy atoms of the homologous template compound, whereas in the second, the shape is additionally annotated with pharmacophore data for some or all beads. For both protocols, ambiguous distance restraints are subsequently defined between those beads and the heavy atoms of the ligand to be docked. We have benchmarked the performance of these protocols with a fully unbound version of the widely used DUD-E (Database of Useful Decoys-Enhanced) dataset. In this unbound docking scenario, our template/shape-based docking protocol reaches an overall success rate of 81% when a reliable template can be identified (which was the case for 99 out of 102 complexes in the DUD-E dataset), which is close to the best results reported for bound docking on the DUD-E dataset.

High Impact: The Role of Promiscuous Binding Sites in Polypharmacology.

The literature focuses on drug promiscuity, which is a drug's ability to bind to several targets, because it plays an essential role in polypharmacology. However, little work has been completed regarding binding site promiscuity, even though its properties are now recognized among the key factors that impact drug promiscuity. Here, we quantified and characterized the promiscuity of druggable binding sites from protein-ligand complexes in the high quality Mother Of All Databases while using statistical methods. Most of the sites (80%) exhibited promiscuity, irrespective of the protein class. Nearly half were highly promiscuous and able to interact with various types of ligands. The corresponding pockets were rather large and hydrophobic, with high sulfur atom and aliphatic residue frequencies, but few side chain atoms. Consequently, their interacting ligands can be large, rigid, and weakly hydrophilic. The selective sites that interacted with one ligand type presented less favorable pocket properties for establishing ligand contacts. Thus, their ligands were highly adaptable, small, and hydrophilic. In the dataset, the promiscuity of the site rather than the drug mainly explains the multiple interactions between the drug and target, as most ligand types are dedicated to one site. This underlines the essential contribution of binding site promiscuity to drug promiscuity between different protein classes.

Predicting the affinity of Farnesoid X Receptor ligands through a hierarchical ranking protocol: a D3R Grand Challenge 2 case study.

The Drug Design Data Resource (D3R) Grand Challenges are blind contests organized to assess the state-of-the-art methods accuracy in predicting binding modes and relative binding free energies of experimentally validated ligands for a given target. The second stage of the D3R Grand Challenge 2 (GC2) was focused on ranking 102 compounds according to their predicted affinity for Farnesoid X Receptor. In this task, our workflow was ranked 5th out of the 77 submissions in the structure-based category. Our strategy consisted in (1) a combination of molecular docking using AutoDock 4.2 and manual edition of available structures for binding poses generation using SeeSAR, (2) the use of HYDE scoring for pose selection, and (3) a hierarchical ranking using HYDE and MM/GBSA. In this report, we detail our pose generation and ligands ranking protocols and provide guidelines to be used in a prospective computer aided drug design program.

Review of in silico studies dedicated to the nuclear receptor family: Therapeutic prospects and toxicological concerns.

Being in the center of both therapeutic and toxicological concerns, NRs are widely studied for drug discovery application but also to unravel the potential toxicity of environmental compounds such as pesticides, cosmetics or additives. High throughput screening campaigns (HTS) are largely used to detect compounds able to interact with this protein family for both therapeutic and toxicological purposes. These methods lead to a large amount of data requiring the use of computational approaches for a robust and correct analysis and interpretation. The output data can be used to build predictive models to forecast the behavior of new chemicals based on their in vitro activities. This atrticle is a review of the studies published in the last decade and dedicated to NR ligands in silico prediction for both therapeutic and toxicological purposes. Over 100 articles concerning 14 NR subfamilies were carefully read and analyzed in order to retrieve the most commonly used computational methods to develop predictive models, to retrieve the databases deployed in the model building process and to pinpoint some of the limitations they faced.

DeepRank: a deep learning framework for data mining 3D protein-protein interfaces.

Three-dimensional (3D) structures of protein complexes provide fundamental information to decipher biological processes at the molecular scale. The vast amount of experimentally and computationally resolved protein-protein interfaces (PPIs) offers the possibility of training deep learning models to aid the predictions of their biological relevance. We present here DeepRank, a general, configurable deep learning framework for data mining PPIs using 3D convolutional neural networks (CNNs). DeepRank maps features of PPIs onto 3D grids and trains a user-specified CNN on these 3D grids. DeepRank allows for efficient training of 3D CNNs with data sets containing millions of PPIs and supports both classification and regression. We demonstrate the performance of DeepRank on two distinct challenges: The classification of biological versus crystallographic PPIs, and the ranking of docking models. For both problems DeepRank is competitive with, or outperforms, state-of-the-art methods, demonstrating the versatility of the framework for research in structural biology.

Nuclear Receptors Database Including Negative Data (NR-DBIND): A Database Dedicated to Nuclear Receptors Binding Data Including Negative Data and Pharmacological Profile.

Nuclear receptors (NRs) are transcription factors that regulate gene expression in various physiological processes through their interactions with small hydrophobic molecules. They constitute an important class of targets for drugs and endocrine disruptors and are widely studied for both health and environment concerns. Since the integration of negative data can be critical for accurate modeling of ligand activity profiles, we manually collected and annotated NRs interaction data (positive and negative) through a sharp review of the corresponding literature. 15 116 positive and negative interactions data are provided for 28 NRs together with 593 PDB structures in the freely available Nuclear Receptors Database Including Negative Data ( http://nr-dbind.drugdesign.fr ). The NR-DBIND contains the most extensive information about interaction data on NRs, which should bring valuable information to chemists, biologists, pharmacologists and toxicologists.

Hits Discovery on the Androgen Receptor: In Silico Approaches to Identify Agonist Compounds.

The androgen receptor (AR) is a transcription factor that plays a key role in sexual phenotype and neuromuscular development. AR can be modulated by exogenous compounds such as pharmaceuticals or chemicals present in the environment, and particularly by AR agonist compounds that mimic the action of endogenous agonist ligands and whether restore or alter the AR endocrine system functions. The activation of AR must be correctly balanced and identifying potent AR agonist compounds is of high interest to both propose treatments for certain diseases, or to predict the risk related to agonist chemicals exposure. The development of in silico approaches and the publication of structural, affinity and activity data provide a good framework to develop rational AR hits prediction models. Herein, we present a docking and a pharmacophore modeling strategy to help identifying AR agonist compounds. All models were trained on the NR-DBIND that provides high quality binding data on AR and tested on AR-agonist activity assays from the Tox21 initiative. Both methods display high performance on the NR-DBIND set and could serve as starting point for biologists and toxicologists. Yet, the pharmacophore models still need data feeding to be used as large scope undesired effect prediction models.

Decoys Selection in Benchmarking Datasets: Overview and Perspectives.

Virtual Screening (VS) is designed to prospectively help identifying potential hits, i.e., compounds capable of interacting with a given target and potentially modulate its activity, out of large compound collections. Among the variety of methodologies, it is crucial to select the protocol that is the most adapted to the query/target system under study and that yields the most reliable output. To this aim, the performance of VS methods is commonly evaluated and compared by computing their ability to retrieve active compounds in benchmarking datasets. The benchmarking datasets contain a subset of known active compounds together with a subset of decoys, i.e., assumed non-active molecules. The composition of both the active and the decoy compounds subsets is critical to limit the biases in the evaluation of the VS methods. In this review, we focus on the selection of decoy compounds that has considerably changed over the years, from randomly selected compounds to highly customized or experimentally validated negative compounds. We first outline the evolution of decoys selection in benchmarking databases as well as current benchmarking databases that tend to minimize the introduction of biases, and secondly, we propose recommendations for the selection and the design of benchmarking datasets.

Discriminating Agonist from Antagonist Ligands of the Nuclear Receptors Using Different Chemoinformatics Approaches.

Nuclear receptors (NRs) constitute an important class of therapeutic targets. During the last 4 years, we tackled the pharmacological profile assessment of NR ligands for which we constructed the NRLiSt BDB. We evaluated and compared the performance of different virtual screening approaches: mean of molecular descriptor distribution values, molecular docking and 3D pharmacophore models. The simple comparison of the distribution profiles of 4885 molecular descriptors between the agonist and antagonist datasets didn't provide satisfying results. We obtained an overall good performance with the docking method we used, Surflex-Dock which was able to discriminate agonist from antagonist ligands. But the availability of PDB structures in the "pharmacological-profile-to-predict-bound-state" (agonist-bound or antagonist-bound) and the availability of enough ligands of both pharmacological profiles constituted limits to generalize this protocol for all NRs. Finally, the 3D pharmacophore modeling approach, allowed us to generate selective agonist pharmacophores and selective antagonist pharmacophores that covered more than 99 % of the whole NRLiSt BDB. This study allowed a better understanding of the pharmacological modulation of NRs with small molecules and could be extended to other therapeutic classes.