Gerard Martínez-Rosell

KDEEP: Protein–Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks

Accurately predicting protein-ligand binding affinities is an important problem in computational chemistry since it can substantially accelerate drug discovery for virtual screening and lead optimization. We propose here a fast machine-learning approach for predicting binding affinities using state-of-the-art 3D-convolutional neural networks and compare this approach to other machine-learning and scoring methods using several diverse data sets. The results for the standard PDBbind (v.2016) core test-set are state-of-the-art with a Pearsons correlation coefficient of 0.82 and a RMSE of 1.27 in pK units between experimental and predicted affinity, but accuracy is still very sensitive to the specific protein used. KDEEP is made available via PlayMolecule.org for users to test easily their own protein-ligand complexes, with each prediction taking a fraction of a second. We believe that the speed, performance, and ease of use of KDEEP makes it already an attractive scoring function for modern computational chemistry pipelines.

DeepSite: protein-binding site predictor using 3D-convolutional neural networks.

Motivation: An important step in structure‐based drug design consists in the prediction of druggable binding sites. Several algorithms for detecting binding cavities, those likely to bind to a small drug compound, have been developed over the years by clever exploitation of geometric, chemical and evolutionary features of the protein. Results: Here we present a novel knowledge‐based approach that uses state‐of‐the‐art convolutional neural networks, where the algorithm is learned by examples. In total, 7622 proteins from the scPDB database of binding sites have been evaluated using both a distance and a volumetric overlap approach. Our machine‐learning based method demonstrates superior performance to two other competitive algorithmic strategies. Availability and implementation: DeepSite is freely available at www.playmolecule.org. Users can submit either a PDB ID or PDB file for pocket detection to our NVIDIA GPU‐equipped servers through a WebGL graphical interface. Contact: gianni.defabritiis@upf.edu Supplementary information: Supplementary data are available at Bioinformatics online.

Drug Discovery and Molecular Dynamics: Methods, Applications and Perspective Beyond the Second Timescale

Bio-molecular dynamics (MD) simulations based on graphical processing units (GPUs) were first released to the public in the early 2009 with the code ACEMD. Almost 8 years after, applications now encompass a broad range of molecular studies, while throughput improvements have opened the way to millisecond sampling timescales. Based on an extrapolation of the amount of sampling in published literature, the second timescale will be reached by the year 2022, and therefore we predict that molecular dynamics is going to become one of the main tools in drug discovery in both academia and industry. Here, we review successful applications in the drug discovery domain developed over these recent years of GPU-based MD. We also retrospectively analyse limitations that have been overcome over the years and give a perspective on challenges that remain to be addressed.

Dynamic and Kinetic Elements of µ-Opioid Receptor Functional Selectivity

While the therapeutic effect of opioids analgesics is mainly attributed to µ-opioid receptor (MOR) activation leading to G protein signaling, their side effects have mostly been linked to β-arrestin signaling. To shed light on the dynamic and kinetic elements underlying MOR functional selectivity, we carried out close to half millisecond high-throughput molecular dynamics simulations of MOR bound to a classical opioid drug (morphine) or a potent G protein-biased agonist (TRV-130). Statistical analyses of Markov state models built using this large simulation dataset combined with information theory enabled, for the first time: a) Identification of four distinct metastable regions along the activation pathway, b) Kinetic evidence of a different dynamic behavior of the receptor bound to a classical or G protein-biased opioid agonist, c) Identification of kinetically distinct conformational states to be used for the rational design of functionally selective ligands that may eventually be developed into improved drugs; d) Characterization of multiple activation/deactivation pathways of MOR, and e) Suggestion from calculated transition timescales that MOR conformational changes are not the rate-limiting step in receptor activation.

Molecular-Simulation-Driven Fragment Screening for the Discovery of New CXCL12 Inhibitors

Fragment-based drug discovery (FBDD) has become a mainstream approach in drug design because it allows the reduction of the chemical space and screening libraries while identifying fragments with high protein-ligand efficiency interactions that can later be grown into drug-like leads. In this work, we leverage high-throughput molecular dynamics (MD) simulations to screen a library of 129 fragments for a total of 5.85 ms against the CXCL12 monomer, a chemokine involved in inflammation and diseases such as cancer. Our in silico binding assay was able to recover binding poses, affinities, and kinetics for the selected library and was able to predict 8 mM-affinity fragments with ligand efficiencies higher than 0.3. All of the fragment hits present a similar chemical structure, with a hydrophobic core and a positively charged group, and bind to either sY7 or H1S68 pockets, where they share pharmacophoric properties with experimentally resolved natural binders. This work presents a large-scale screening assay using an exclusive combination of thousands of short MD adaptive simulations analyzed with a Markov state model (MSM) framework.

PlayMolecule ProteinPrepare: A Web Application for Protein Preparation for Molecular Dynamics Simulations

Protein preparation is a critical step in molecular simulations that consists of refining a Protein Data Bank (PDB) structure by assigning titration states and optimizing the hydrogen-bonding network. In this application note, we describe ProteinPrepare, a web application designed to interactively support the preparation of protein structures. Users can upload a PDB file, choose the solvent pH value, and inspect the resulting protonated residues and hydrogen-bonding network within a 3D web interface. Protonation states are suggested automatically but can be manually changed using the visual aid of the hydrogen-bonding network. Tables and diagrams provide estimated pKa values and charge states, with visual indication for cases where review is required. We expect the graphical interface to be a useful instrument to assess the validity of the preparation, but nevertheless, a script to execute the preparation offline with the High-Throughput Molecular Dynamics (HTMD) environment is also provided for noninteractive operations.

Simulations meet machine learning in structural biology

Classical molecular dynamics (MD) simulations will be able to reach sampling in the second timescale within five years, producing petabytes of simulation data at current force field accuracy. Notwithstanding this, MD will still be in the regime of low-throughput, high-latency predictions with average accuracy. We envisage that machine learning (ML) will be able to solve both the accuracy and time-to-prediction problem by learning predictive models using expensive simulation data. The synergies between classical, quantum simulations and ML methods, such as artificial neural networks, have the potential to drastically reshape the way we make predictions in computational structural biology and drug discovery.

High-Throughput Automated Preparation and Simulation of Membrane Proteins with HTMD

HTMD is a programmable scientific platform intended to facilitate simulation-based research in molecular systems. This paper presents the functionalities of HTMD for the preparation of a molecular dynamics simulation starting from PDB structures, building the system using well-known force fields, and applying standardized protocols for running the simulations. We demonstrate the frameworks flexibility for high-throughput molecular simulations by applying a preparation, building, and simulation protocol with multiple force-fields on all of the seven hundred eukaryotic membrane proteins resolved to-date from the orientation of proteins in membranes (OPM) database. All of the systems are available on www.playmolecule.org .

PlayMolecule BindScope: large scale CNN-based virtual screening on the web

SUMMARY Virtual screening pipelines are one of the most popular used tools in structure-based drug discovery, since they can can reduce both time and cost associated with experimental assays. Recent advances in deep learning methodologies have shown that these outperform classical scoring functions at discriminating binder protein-ligand complexes. Here, we present BindScope, a web application for large-scale active-inactive classification of compounds based on deep convolutional neural networks. Performance is on a pair with current state-of-the-art pipelines. Users can screen on the order of hundreds of compounds at once and interactively visualize the results. AVAILABILITY AND IMPLEMENTATION BindScope is available as part of the PlayMolecule.org web application suite. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

LigVoxel: inpainting binding pockets using 3D-convolutional neural networks

Motivation: Structure‐based drug discovery methods exploit protein structural information to design small molecules binding to given protein pockets. This work proposes a purely data driven, structure‐based approach for imaging ligands as spatial fields in target protein pockets. We use an end‐to‐end deep learning framework trained on experimental protein‐ligand complexes with the intention of mimicking a chemists intuition at manually placing atoms when designing a new compound. We show that these models can generate spatial images of ligand chemical properties like occupancy, aromaticity and donor‐acceptor matching the protein pocket. Results: The predicted fields considerably overlap with those of unseen ligands bound to the target pocket. Maximization of the overlap between the predicted fields and a given ligand on the Astex diverse set recovers the original ligand crystal poses in 70 out of 85 cases within a threshold of 2 Å RMSD. We expect that these models can be used for guiding structure‐based drug discovery approaches. Availability and implementation: LigVoxel is available as part of the PlayMolecule.org molecular web application suite. Supplementary information: Supplementary data are available at Bioinformatics online.