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.

iScore: a novel graph kernel-based function for scoring protein-protein docking models.

MOTIVATION: Protein complexes play critical roles in many aspects of biological functions. Three-dimensional (3D) structures of protein complexes are critical for gaining insights into structural bases of interactions and their roles in the biomolecular pathways that orchestrate key cellular processes. Because of the expense and effort associated with experimental determinations of 3D protein complex structures, computational docking has evolved as a valuable tool to predict 3D structures of biomolecular complexes. Despite recent progress, reliably distinguishing near-native docking conformations from a large number of candidate conformations, the so-called scoring problem, remains a major challenge. RESULTS: Here we present iScore, a novel approach to scoring docked conformations that combines HADDOCK energy terms with a score obtained using a graph representation of the protein-protein interfaces and a measure of evolutionary conservation. It achieves a scoring performance competitive with, or superior to, that of state-of-the-art scoring functions on two independent datasets: (i) Docking software-specific models and (ii) the CAPRI score set generated by a wide variety of docking approaches (i.e. docking software-non-specific). iScore ranks among the top scoring approaches on the CAPRI score set (13 targets) when compared with the 37 scoring groups in CAPRI. The results demonstrate the utility of combining evolutionary, topological and energetic information for scoring docked conformations. This work represents the first successful demonstration of graph kernels to protein interfaces for effective discrimination of near-native and non-native conformations of protein complexes. AVAILABILITY AND IMPLEMENTATION: The iScore code is freely available from Github: https://github.com/DeepRank/iScore (DOI: 10.5281/zenodo.2630567). And the docking models used are available from SBGrid: https://data.sbgrid.org/dataset/684). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The effect of the magnitude and direction of the dipoles of organic cations on the electronic structure of hybrid halide perovskites.

We present ab initio calculations (DFT and SOC-G0W0) of the optoelectronic properties of different hybrid-halide perovskites, namely X-PbI3 (X = methylamonimum, formamidinium, guanidinium, hydrazinium, and hydroxylammonium). These calculations shed new light on how the substitution of different organic cations in the material influences its optoelectronic properties. Our simulations show a significant modification of the lattice parameter and band gap of the material upon cation substitution. These modifications are not only due to steric effects but also due to electrostatic interactions between the organic and inorganic parts of the material. In addition to this, we demonstrate how the relative orientations of neighboring cations in the material modify the local electrostatic potential of the system and its fundamental band gap. This change in the band gap is accompanied by the formation of localized and spatially separated electronic states. These localized states modify the carrier mobility in the materials and can be a reason for the formation and recombination of the charge carriers in these very promising materials.

High Electronic Conductance through Double-Helix DNA Molecules with Fullerene Anchoring Groups.

Determining the mechanism of charge transport through native DNA remains a challenge as different factors such as measuring conditions, molecule conformations, and choice of technique can significantly affect the final results. In this contribution, we have used a new approach to measure current flowing through isolated double-stranded DNA molecules, using fullerene groups to anchor the DNA to a gold substrate. Measurements were performed at room temperature in an inert environment using a conductive AFM technique. It is shown that the π-stacked B-DNA structure is conserved on depositing the DNA. As a result, currents in the nanoampere range were obtained for voltages ranging between ±1 V. These experimental results are supported by a theoretical model that suggests that a multistep hopping mechanism between delocalized domains is responsible for the long-range current flow through this specific type of DNA.

Computational design of donor-bridge-acceptor systems exhibiting pronounced quantum interference effects.

Quantum interference is a well-known phenomenon that dictates charge transport properties of single molecule junctions. However, reports on quantum interference in donor-bridge-acceptor molecules are scarce. This might be due to the difficulties in meeting the conditions for the presence of quantum interference in a donor-bridge-acceptor system. The electronic coupling between the donor, bridge, and acceptor moieties must be weak in order to ensure localised initial and final states for charge transfer. Yet, it must be strong enough to allow all bridge orbitals to mediate charge transfer. We present the computational route to the design of a donor-bridge-acceptor molecule that features the right balance between these contradicting requirements and exhibits pronounced interference effects.

Impact of a single base pair substitution on the charge transfer rate along short DNA hairpins.

Numerical studies of hole migration along short DNA hairpins were performed with a particular emphasis on the variations of the rate and quantum yield of the charge separation process with the location of a single guanine:cytosine (G:C) base pair. Our calculations show that the hole arrival rate increases as the position of the guanine:cytosine base pair shifts from the beginning to the end of the sequence. Although these results are in agreement with recent experimental findings, the mechanism governing the charge migration along these sequences is revisited here. Instead of the phenomenological two-step hopping mechanism via the guanine base, the charge propagation occurs through a delocalization of the hole density along the base pair stack. Furthermore, the variations of the charge transfer with the position of the guanine base are explained by the impact of the base pair substitutions on the delocalized conduction channels.

Charge Transport across DNA-Based Three-Way Junctions.

DNA-based molecular electronics will require charges to be transported from one site within a 2D or 3D architecture to another. While this has been shown previously in linear, π-stacked DNA sequences, the dynamics and efficiency of charge transport across DNA three-way junction (3WJ) have yet to be determined. Here, we present an investigation of hole transport and trapping across a DNA-based three-way junction systems by a combination of femtosecond transient absorption spectroscopy and molecular dynamics simulations. Hole transport across the junction is proposed to be gated by conformational fluctuations in the ground state which bring the transiently populated hole carrier nucleobases into better aligned geometries on the nanosecond time scale, thus modulating the π-π electronic coupling along the base pair sequence.

Different mechanisms for hole and electron transfer along identical molecular bridges: the importance of the initial state delocalization.

We report measurements of hole and electron transfer along identical oligo-p-phenylene molecular bridges of increasing length. Although the injection barriers for hole and electron transfer are similar, we observed striking differences in the distance dependence and absolute magnitude of the rates of these two processes. Electron transfer is characterized by an almost distance-independent, fast charge-transfer rate. Hole transfer presents a much slower rate that decreases significantly with the length of the bridge. Time-dependent density functional calculations show that the observed differences can be explained by the delocalization of the respective initial excitation. The evaluation of the initial state is therefore essential when comparing charge-transfer rates between different donor-bridge-acceptor systems.

Between superexchange and hopping: an intermediate charge-transfer mechanism in poly(A)-poly(T) DNA hairpins.

We developed a model for hole migration along relatively short DNA hairpins with fewer that seven adenine (A):thymine (T) base pairs. The model was used to simulate hole migration along poly(A)-poly(T) sequences with a particular emphasis on the impact of partial hole localization on the different rate processes. The simulations, performed within the framework of the stochastic surrogate Hamiltonian approach, give values for the arrival rate in good agreement with experimental data. Theoretical results obtained for hairpins with fewer than three A:T base pairs suggest that hole transfer along short hairpins occurs via superexchange. This mechanism is characterized by the exponential distance dependence of the arrival rate on the donor/acceptor distance, k(a) ≃ e(-ßR), with ß = 0.9 Å(-1). For longer systems, up to six A:T pairs, the distance dependence follows a power law k(a) ≃ R(-η) with η = 2. Despite this seemingly clear signature of unbiased hopping, our simulations show the complete delocalization of the hole density along the entire hairpin. According to our analysis, the hole transfer along relatively long sequences may proceed through a mechanism which is distinct from both coherent single-step superexchange and incoherent multistep hopping. The criterion for the validity of this mechanism intermediate between superexchange and hopping is proposed. The impact of partial localization on the rate of hole transfer between neighboring A bases was also investigated.

CHROMagar Yersinia, a new chromogenic agar for screening of potentially pathogenic Yersinia enterocolitica isolates in stools.

CHROMagar Yersinia (CAY) is a new chromogenic medium for the presumptive detection of virulent Yersinia enterocolitica in stools. Based on a comparative analysis of 1,494 consecutive stools from hospitalized patients, CAY was found to be just as sensitive as the reference medium (cefsulodin-irgasan-novobiocin agar) but was significantly more specific and had a very low false-positive rate. CAY reduces the workload (and thus costs) for stool analysis and can therefore be recommended for routine laboratory use.

Quantum interferences and electron transfer in photosystem I.

We have studied the electron transfer occurring in the photosystem I (PSI) reaction center from the special pair to the first iron-sulfur cluster. Electronic structure calculations performed at the DFT level were employed to determine the on-site energies of the fragments comprising PSI, as well as the charge transfer integrals between neighboring pairs. This electronic Hamiltonian was then used to compute the charge transfer dynamics, using the stochastic surrogate Hamiltonian approach to account for the coherent propagation of the electronic density but also for its energy relaxation and decoherence. These simulations give reasonable transfer time ranging from subpicoseconds to nanoseconds and predict coherent oscillations for several picoseconds. Due to these long-lasting coherences, the propagation of the electronic density can be enhanced or inhibited by quantum interferences. The impact of random fluctuations and asymmetries on these interferences is then discussed. Random fluctuations lead to a classical transport where both constructive and destructive quantum interferences are suppressed. Finally it is shown that an energy difference of 0.15 eV between the on-site energies of the phylloquinones leads to a highly efficient electron transfer even in presence of strong random fluctuations.

Antimicrobial-Resistant Bacteria Carriage in Rodents According to Habitat Anthropization.

It is increasingly suggested that the dynamics of antimicrobial-resistant bacteria in the wild are mostly anthropogenically driven, but the spatial and temporal scales at which these phenomena occur in landscapes are only partially understood. Here, we explore this topic by studying antimicrobial resistance in the commensal bacteria from micromammals sampled at 12 sites from a large heterogenous landscape (the Carmargue area, Rhone Delta) along a gradient of anthropization: natural reserves, rural areas, towns, and sewage-water treatment plants. There was a positive relationship between the frequency of antimicrobial-resistant bacteria and the level of habitat anthropization. Although low, antimicrobial resistance was also present in natural reserves, even in the oldest one, founded in 1954. This study is one of the first to support the idea that rodents in human-altered habitats are important components of the environmental pool of resistance to clinically relevant antimicrobials and also that a "One Health" approach is required to assess issues related to antimicrobial resistance dynamics in anthropized landscapes.

Single molecule logical devices.

After almost 40 years of development, molecular electronics has given birth to many exciting ideas that range from molecular wires to molecular qubit-based quantum computers. This chapter reviews our efforts to answer a simple question: how smart can a single molecule be? In our case a molecule able to perform a simple Boolean function is a child prodigy. Following the Aviram and Ratner approach, these molecules are inserted between several conducting electrodes. The electronic conduction of the resulting molecular junction is extremely sensitive to the chemical nature of the molecule. Therefore designing this latter correctly allows the implementation of a given function inside the molecular junction. Throughout the chapter different approaches are reviewed, from hybrid devices to quantum molecular logic gates. We particularly stress that one can implement an entire logic circuit in a single molecule, using either classical-like intramolecular connections, or a deformation of the molecular orbitals induced by a conformational change of the molecule. These approaches are radically different from the hybrid-device approach, where several molecules are connected together to build the circuit.

PANDORA: A Fast, Anchor-Restrained Modelling Protocol for Peptide: MHC Complexes.

Deeper understanding of T-cell-mediated adaptive immune responses is important for the design of cancer immunotherapies and antiviral vaccines against pandemic outbreaks. T-cells are activated when they recognize foreign peptides that are presented on the cell surface by Major Histocompatibility Complexes (MHC), forming peptide:MHC (pMHC) complexes. 3D structures of pMHC complexes provide fundamental insight into T-cell recognition mechanism and aids immunotherapy design. High MHC and peptide diversities necessitate efficient computational modelling to enable whole proteome structural analysis. We developed PANDORA, a generic modelling pipeline for pMHC class I and II (pMHC-I and pMHC-II), and present its performance on pMHC-I here. Given a query, PANDORA searches for structural templates in its extensive database and then applies anchor restraints to the modelling process. This restrained energy minimization ensures one of the fastest pMHC modelling pipelines so far. On a set of 835 pMHC-I complexes over 78 MHC types, PANDORA generated models with a median RMSD of 0.70 Å and achieved a 93% success rate in top 10 models. PANDORA performs competitively with three pMHC-I modelling state-of-the-art approaches and outperforms AlphaFold2 in terms of accuracy while being superior to it in speed. PANDORA is a modularized and user-configurable python package with easy installation. We envision PANDORA to fuel deep learning algorithms with large-scale high-quality 3D models to tackle long-standing immunology challenges.

Multiresistant Enterobacteriaceae in yellow-legged gull chicks in their first weeks of life.

Wild animal species living in anthropogenic areas are commonly carriers of antimicrobial-resistant bacteria (AMRB), but their role in the epidemiology of these bacteria is unclear. Several studies on AMRB in wildlife have been cross-sectional in design and sampled individual animals at only one point in time. To further understand the role of wildlife in maintaining and potentially transmitting these bacteria to humans and livestock, longitudinal studies are needed in which samples are collected from individual animals over multiple time periods. In Europe, free-ranging yellow-legged gulls (Larus michahellis) commonly live in industrialized areas, forage in landfills, and have been found to carry AMRB in their feces. Using bacterial metagenomics and antimicrobial resistance characterization, we investigated the spatial and temporal patterns of AMRB in a nesting colony of yellow-legged gulls from an industrialized area in southern France. We collected 54 cloacal swabs from 31 yellow-legged gull chicks in 20 nests on three dates in 2016. We found that AMRB in chicks increased over time and was not spatially structured within the gull colony. This study highlights the complex occurrence of AMRB in a free-ranging wildlife species and contributes to our understanding of the public health risks and implications associated with ARMB-carrying gulls living in anthropogenic areas.

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.

Embolus Retriever with Interlinked Cages (ERIC) versus conventional stent retrievers for thrombectomy: a propensity score-based analysis.

BACKGROUND: The Embolus Retriever with Interlinked Cages (ERIC) is one of the latest devices for thrombectomies. It has several architectural features that are supposed to enhance its ability to remove clots and prevent distal emboli. We aimed to compare ERIC with standard stent retrievers (SRs) using propensity score (PS) matching. METHODS: The clinical and radiological data of all consecutive patients treated with ERIC or standard FDA-approved stent retrievers were collected from a prospective multicenter registry. We compared procedural outcomes (recanalization rates according to the modified Thrombolysis In Cerebral Infarction (mTICI) score and procedural complications) and clinical outcomes (modified Rankin Scale (mRS) and mortality at 3 months). Matching of the populations with PS was performed to account for differences in baseline characteristics. RESULTS: A total of 1230 patients were included. In both the PS-matched cohort (195 ERIC patients, 630 SR patients) and the inverse probability of treatment weighting PS-adjusted cohort (206 ERIC patients, 1024 SR patients) there was no difference in terms of successful recanalization (modified TICI score ≥2b), good clinical outcome (mRS=0-2 or equal to pre-stroke mRS), or mortality at 3 months. Patients treated with first-line ERIC had a higher rate of complete recanalization (mTICI 3); however, they also required more passes and more frequent rescue therapy than the SR patient group. CONCLUSION: In a large multicenter registry with PS matching, the ERIC device provided equivalent angiographic and clinical results to conventional SRs. CLINICAL TRIAL REGISTRATION: URL: http://www.clinicaltrials.gov Unique identifier: NCT03776877.

iScore: An MPI supported software for ranking protein-protein docking models based on a random walk graph kernel and support vector machines.

Computational docking is a promising tool to model three-dimensional (3D) structures of protein-protein complexes, which provides fundamental insights of protein functions in the cellular life. Singling out near-native models from the huge pool of generated docking models (referred to as the scoring problem) remains as a major challenge in computational docking. We recently published iScore, a novel graph kernel based scoring function. iScore ranks docking models based on their interface graph similarities to the training interface graph set. iScore uses a support vector machine approach with random-walk graph kernels to classify and rank protein-protein interfaces. Here, we present the software for iScore. The software provides executable scripts that fully automate the computational workflow. In addition, the creation and analysis of the interface graph can be distributed across different processes using Message Passing interface (MPI) and can be offloaded to GPUs thanks to dedicated CUDA kernels.

Signatures of Conformational Dynamics and Electrode-Molecule Interactions in the Conductance Profile During Pulling of Single-Molecule Junctions.

We demonstrate that conductance can act as a sensitive probe of conformational dynamics and electrode-molecule interactions during the equilibrium and nonequilibrium pulling of molecular junctions. To do so, we use a combination of classical molecular dynamics simulations and Landauer electron transport computations to investigate the conductance of a family of Au-alkanedithiol-Au junctions as they are mechanically elongated. The simulations show an overall decay of the conductance during pulling that is due to a decrease in the through-space electrode-molecule interactions, and that sensitivity depends on the electrode geometry. In addition, characteristic kinks induced by level alignment shifts (and to a lesser extent by quantum destructive interference) were also observed superimposed to the overall decay during pulling simulations. The latter effect depends on the variation of the molecular dihedral angles during pulling and therefore offers an efficient solution to experimentally monitor conformational dynamics at the single-molecule limit.

Computational Design of Two-Dimensional Perovskites with Functional Organic Cations.

Two-dimensional (2D) halide perovskites are a class of materials in which 2D layers of perovskite are separated by large organic cations. Conventionally, the 2D perovskites incorporate organic cations as spacers, but these organic cations also offer a route to introduce specific functionality in the material. In this work, we demonstrate, by density functional theory calculations, that the introduction of electron withdrawing and electron donating molecules leads to the formation of localized states, either in the organic or the inorganic part. Furthermore, we show that the energy of the bands located in the organic and inorganic parts can be tuned independently. The organic cation levels can be tuned by changing the electron withdrawing/donating character, whereas the energy levels in the inorganic part can be modified by varying the number of inorganic perovskite layers. This opens a new window for the design of 2D perovskites with properties tuned for specific applications.