A Perspective on the Prospective Use of AI in Protein Structure Prediction.

AlphaFold2 (AF2) and RoseTTaFold (RF) have revolutionized structural biology, serving as highly reliable and effective methods for predicting protein structures. This article explores their impact and limitations, focusing on their integration into experimental pipelines and their application in diverse protein classes, including membrane proteins, intrinsically disordered proteins (IDPs), and oligomers. In experimental pipelines, AF2 models help X-ray crystallography in resolving the phase problem, while complementarity with mass spectrometry and NMR data enhances structure determination and protein flexibility prediction. Predicting the structure of membrane proteins remains challenging for both AF2 and RF due to difficulties in capturing conformational ensembles and interactions with the membrane. Improvements in incorporating membrane-specific features and predicting the structural effect of mutations are crucial. For intrinsically disordered proteins, AF2's confidence score (pLDDT) serves as a competitive disorder predictor, but integrative approaches including molecular dynamics (MD) simulations or hydrophobic cluster analyses are advocated for accurate dynamics representation. AF2 and RF show promising results for oligomeric models, outperforming traditional docking methods, with AlphaFold-Multimer showing improved performance. However, some caveats remain in particular for membrane proteins. Real-life examples demonstrate AF2's predictive capabilities in unknown protein structures, but models should be evaluated for their agreement with experimental data. Furthermore, AF2 models can be used complementarily with MD simulations. In this Perspective, we propose a "wish list" for improving deep-learning-based protein folding prediction models, including using experimental data as constraints and modifying models with binding partners or post-translational modifications. Additionally, a meta-tool for ranking and suggesting composite models is suggested, driving future advancements in this rapidly evolving field.

It Takes Tau to Tango: Investigating the Fuzzy Interaction between the R2-Repeat Domain and Tubulin C-Terminal Tails.

The microtubule-associated protein (MAP) tau plays a key role in the regulation of microtubule assembly and spatial organization. Tau hyperphosphorylation affects its binding on the tubulin surface and has been shown to be involved in several pathologies such as Alzheimer's disease. As the tau binding site on the microtubule lays close to the disordered and highly flexible tubulin C-terminal tails (CTTs), these are likely to impact the tau-tubulin interaction. Since the disordered tubulin CTTs are missing from the available experimental structures, we used homology modeling to build two complete models of tubulin heterotrimers with different isotypes for the ß-tubulin subunit (ßI/αI/ßI and ßIII/αI/ßIII). We then performed long timescale classical Molecular Dynamics simulations for the tau-R2/tubulin assembly (in systems with and without CTTs) and analyzed the resulting trajectories to obtain a detailed view of the protein interface in the complex and the impact of the CTTs on the stability of this assembly. Additional analyses of the CTT mobility in the presence, or in the absence, of tau also highlight how tau might modulate the CTT activity as hooks that are involved in the recruitment of several MAPs. In particular, we observe a wrapping phenomenon, where the ß-tubulin CTTs form a loop over tau-R2, thus stabilizing its interaction with the tubulin surface and simultaneously reducing the CTT availability for interactions with other MAPs.

Best Practices of Using AI-Based Models in Crystallography and Their Impact in Structural Biology.

The recent breakthrough made in the field of three-dimensional (3D) structure prediction by artificial intelligence softwares, such as initially AlphaFold2 (AF2) and RosettaFold (RF) and more recently large Language Models (LLM), has revolutionized the field of structural biology in particular and also biology as a whole. These models have clearly generated great enthusiasm within the scientific community, and different applications of these 3D predictions are regularly described in scientific articles, demonstrating the impact of these high-quality models. Despite the acknowledged high accuracy of these models in general, it seems important to make users of these models aware of the wealth of information they offer and to encourage them to make the best use of them. Here, we focus on the impact of these models in a specific application by structural biologists using X-ray crystallography. We propose guidelines to prepare models to be used for molecular replacement trials to solve the phase problem. We also encourage colleagues to share as much detail as possible about how they use these models in their research, where the models did not yield correct molecular replacement solutions, and how these predictions fit with their experimental 3D structure. We feel this is important to improve the pipelines using these models and also to get feedback on their overall quality.

Design - a new way to look at old molecules.

We discuss how design enriches molecular science, particularly structural biology and bioinformatics. We present two use cases, one in academic practice and the other to design for outreach. The first case targets the representation of ion channels and their dynamic properties. In the second, we document a transition process from a research environment to general-purpose designs. Several testimonials from practitioners are given. By describing the design process of abstracted shapes, exploded views of molecular structures, motion-averaged slices, 360-degree panoramic projections, and experiments with lit sphere shading, we document how designers help make scientific data accessible without betraying its meaning, and how a creative mind adds value over purely data-driven visualizations. A similar conclusion was drawn for public outreach, as we found that comic-book-style drawings are better suited for communicating science to a broad audience.

Modeling the Dynamics of Protein-Protein Interfaces, How and Why?

Protein-protein assemblies act as a key component in numerous cellular processes. Their accurate modeling at the atomic level remains a challenge for structural biology. To address this challenge, several docking and a handful of deep learning methodologies focus on modeling protein-protein interfaces. Although the outcome of these methods has been assessed using static reference structures, more and more data point to the fact that the interaction stability and specificity is encoded in the dynamics of these interfaces. Therefore, this dynamics information must be taken into account when modeling and assessing protein interactions at the atomistic scale. Expanding on this, our review initially focuses on the recent computational strategies aiming at investigating protein-protein interfaces in a dynamic fashion using enhanced sampling, multi-scale modeling, and experimental data integration. Then, we discuss how interface dynamics report on the function of protein assemblies in globular complexes, in fuzzy complexes containing intrinsically disordered proteins, as well as in active complexes, where chemical reactions take place across the protein-protein interface.

Between Two Walls: Modeling the Adsorption Behavior of ß-Glucosidase A on Bare and SAM-Functionalized Gold Surfaces.

The efficient immobilization of enzymes on surfaces remains a complex but central issue in the biomaterials field, which requires us to understand this process at the atomic level. Using a multiscale approach combining all-atom molecular dynamics and coarse-grain Brownian dynamics simulations, we investigated the adsorption behavior of ß-glucosidase A (ßGA) on bare and self-assembled monolayer (SAM)-functionalized gold surfaces. We monitored the enzyme position and orientation during the molecular dynamics (MD) trajectories and measured the contacts it forms with both surfaces. While the adsorption process has little impact on the protein conformation, it can nonetheless perturb its mechanical properties and catalytic activity. Our results show that compared to the SAM-functionalized surface, the adsorption of ßGA on bare gold is more stable, but less specific, and more likely to disrupt the enzyme's function. This observation emphasizes the fact that the structural organization of proteins at the solid interface is a key point when designing devices based on enzyme immobilization, as one must find an acceptable stability-activity trade-off.

Conformational and mechanical stability of the isolated large subunit of membrane-bound [NiFe]-hydrogenase from Cupriavidus necator.

Comprising at least a bipartite architecture, the large subunit of [NiFe]-hydrogenase harbors the catalytic nickel-iron site while the small subunit houses an array of electron-transferring Fe-S clusters. Recently, some [NiFe]-hydrogenase large subunits have been isolated showing an intact and redox active catalytic cofactor. In this computational study we have investigated one of these metalloproteins, namely the large subunit HoxG of the membrane-bound hydrogenase from Cupriavidus necator (CnMBH), targeting its conformational and mechanical stability using molecular modelling and long all-atom Gaussian accelerated molecular dynamics (GaMD). Our simulations predict that isolated HoxG is stable in aqueous solution and preserves a large portion of its mechanical properties, but loses rigidity in regions around the active site, in contrast to the MBH heterodimer. Inspired by biochemical data showing dimerization of the HoxG protein and IR measurements revealing an increased stability of the [NiFe] cofactor in protein preparations with higher dimer content, corresponding simulations of homodimeric forms were also undertaken. While the monomeric subunit contains several flexible regions, our data predicts a regained rigidity in homodimer models. Furthermore, we computed the electrostatic properties of models obtained by enhanced sampling with GaMD, which displays a significant amount of positive charge at the protein surface, especially in solvent-exposed former dimer interfaces. These data offer novel insights on the way the [NiFe] core is protected from de-assembly and provide hints for enzyme anchoring to surfaces, which is essential information for further investigations on these minimal enzymes.

When Order Meets Disorder: Modeling and Function of the Protein Interface in Fuzzy Complexes.

The degree of proteins structural organization ranges from highly structured, compact folding to intrinsic disorder, where each degree of self-organization corresponds to specific functions: well-organized structural motifs in enzymes offer a proper environment for precisely positioned functional groups to participate in catalytic reactions; at the other end of the self-organization spectrum, intrinsically disordered proteins act as binding hubs via the formation of multiple, transient and often non-specific interactions. This review focusses on cases where structurally organized proteins or domains associate with highly disordered protein chains, leading to the formation of interfaces with varying degrees of fuzziness. We present a review of the computational methods developed to provide us with information on such fuzzy interfaces, and how they integrate experimental information. The discussion focusses on two specific cases, microtubules and homologous recombination nucleoprotein filaments, where a network of intrinsically disordered tails exerts regulatory function in recruiting partner macromolecules, proteins or DNA and tuning the atomic level association. Notably, we show how computational approaches such as molecular dynamics simulations can bring new knowledge to help bridging the gap between experimental analysis, that mostly concerns ensemble properties, and the behavior of individual disordered protein chains that contribute to regulation functions.

Moving pictures: Reassessing docking experiments with a dynamic view of protein interfaces.

The modeling of protein assemblies at the atomic level remains a central issue in structural biology, as protein interactions play a key role in numerous cellular processes. This problem is traditionally addressed using docking tools, where the quality of the models is based on their similarity to a single reference experimental structure. However, using a static reference does not take into account the dynamic quality of the protein interface. Here, we used all-atom classical Molecular Dynamics simulations to investigate the stability of the reference interface for three complexes that previously served as targets in the CAPRI competition. For each one of these targets, we also ran MD simulations for ten models that are distributed over the High, Medium and Acceptable accuracy categories. To assess the quality of these models from a dynamic perspective, we set up new criteria which take into account the stability of the reference experimental protein interface. We show that, when the protein interfaces are allowed to evolve along time, the original ranking based on the static CAPRI criteria no longer holds as over 50% of the docking models undergo a category change (which can be either toward a better or a lower accuracy group) when reassessing their quality using dynamic information.

Implicit Modeling of the Impact of Adsorption on Solid Surfaces for Protein Mechanics and Activity with a Coarse-Grained Representation.

Surface immobilized enzymes play a key role in numerous biotechnological applications such as biosensors, biofuel cells, or biocatalytic synthesis. As a consequence, the impact of adsorption on the enzyme structure, dynamics, and function needs to be understood on the molecular level as it is critical for the improvement of these technologies. With this perspective in mind, we used a theoretical approach for investigating local protein flexibility on the residue scale that couples a simplified protein representation with an elastic network and Brownian dynamics simulations. The impact of protein adsorption on a solid surface is implicitly modeled via additional external constraints between the residues in contact with the surface. We first performed calculations on a redox enzyme, bilirubin oxidase (BOD) from M. verrucaria, to study the impact of adsorption on its mechanical properties. The resulting rigidity profiles show that, in agreement with the available experimental data, the mechanical variations observed in the adsorbed BOD will depend on its orientation and its anchor residues (i.e., residues that are in contact with the functionalized surface). Additional calculations on ribonuclease A and nitroreductase shed light on how seemingly stable adsorbed enzymes can nonetheless display an important decrease in their catalytic activity resulting from a perturbation of their mechanics and internal dynamics.

Protein Interaction Energy Landscapes are Shaped by Functional and also Non-functional Partners.

In the crowded cell, a strong selective pressure operates on the proteome to limit the competition between functional and non-functional protein-protein interactions. We developed an original theoretical framework in order to interrogate how this competition constrains the behavior of proteins with respect to their partners or random encounters. Our theoretical framework relies on a two-dimensional (2D) representation of interaction energy landscapes, with 2D energy maps, which reflect in a synthetic way the spatial distribution of the interaction propensity of a protein surface for another protein. We realized the interaction propensity mapping of proteins' surfaces in interaction with functional and arbitrary partners and asked whether the distribution of their interaction propensity is conserved during evolution. Therefore, we performed several thousands of cross-docking simulations to systematically characterize the energy landscapes of 103 proteins interacting with different sets of homologs, corresponding to their functional partner's family or arbitrary protein families. Then, we systematically compared the energy maps resulting from the docking of each protein with the different protein families of the dataset. Strikingly, we show that the interaction propensity not only of the binding sites but also of the rest of the surface is conserved for docking partners belonging to the same protein family. Interestingly, this observation holds for docked proteins corresponding to true but also arbitrary partners. Our theoretical framework enables the characterization of the energy behavior of a protein in interaction with hundreds of proteins and opens the way for the characterization of the behavior of proteins in a specific environment.

Coarse-grain simulations on NMR conformational ensembles highlight functional residues in proteins.

Dynamics are a key feature of protein function, and this is especially true of gating residues, which occupy cavity or tunnel lining positions in the protein structure, and will reversibly switch between open and closed conformations in order to control the diffusion of small molecules within a protein's internal matrix. Earlier work on globins and hydrogenases have shown that these gating residues can be detected using a multiscale scheme combining all-atom classic molecular dynamics simulations and coarse-grain calculations of the resulting conformational ensemble mechanical properties. Here, we show that the structural variations observed in the conformational ensembles produced by NMR spectroscopy experiments are sufficient to induce noticeable mechanical changes in a protein, which in turn can be used to identify residues important for function and forming a mechanical nucleus in the protein core. This new approach, which combines experimental data and rapid coarse-grain calculations and no longer needs to resort to time-consuming all-atom simulations, was successfully applied to five different protein families.

Conformational Stability Adaptation of a Double-Stranded RNA-Binding Domain to Transfer RNA Ligand.

The double-stranded RNA-binding domain (dsRBD) is a broadly distributed domain among RNA-maturing enzymes. Although this domain recognizes dsRNA's structures via a conserved canonical structure adopting an α1-ß1ß2ß3-α2 topology, several dsRBDs can accommodate discrete structural extensions expanding further their functional repertoire. How these structural elements engage cooperative communications with the canonical structure and how they contribute to the dsRBD's overall folding are poorly understood. Here, we addressed these issues using the dsRBD of human dihydrouridine synthase-2 (hDus2) (hDus2-dsRBD) as a model. This dsRBD harbors N- and C-terminal extensions, the former being directly involved in the recognition of tRNA substrate of hDus2. These extensions engage residues that form a long-range hydrophobic network (LHN) outside the RNA-binding interface. We show by coarse-grain Brownian dynamics that the Nt-extension and its residues F359 and Y364 rigidify the major folding nucleus of the canonical structure via an indirect effect. hDus2-dsRBD unfolds following a two-state cooperative model, whereas both F359A and Y364A mutants, designed to destabilize this LHN, unfold irreversibly. Structural and computational analyses show that these mutants are unstable due to an increase in the dynamics of the two extensions favoring solvent exposure of α2-helix and weakening the main folding nucleus rigidity. This LHN appears essential for maintaining a thermodynamic stability of the overall system and eventually a functional conformation for tRNA recognition. Altogether, our findings suggest that functional adaptability of extended dsRBDs is promoted by a cooperative hydrophobic coupling between the extensions acting as effectors and the folding nucleus of the canonical structure.

Mechanical variations in proteins with large-scale motions highlight the formation of structural locks.

Protein function depends just as much on flexibility as on structure, and in numerous cases, a protein's biological activity involves transitions that will impact both its conformation and its mechanical properties. Here, we use a coarse-grain approach to investigate the impact of structural changes on protein flexibility. More particularly, we focus our study on proteins presenting large-scale motions. We show how calculating directional force constants within residue pairs, and investigating their variation upon protein closure, can lead to the detection of a limited set of residues that form a structural lock in the protein's closed conformation. This lock, which is composed of residues whose side-chains are tightly interacting, highlights a new class of residues that are important for protein function by stabilizing the closed structure, and that cannot be detected using earlier tools like local rigidity profiles or distance variations maps, or alternative bioinformatics approaches, such as coevolution scores.

Hidden partners: Using cross-docking calculations to predict binding sites for proteins with multiple interactions.

Protein-protein interactions control a large range of biological processes and their identification is essential to understand the underlying biological mechanisms. To complement experimental approaches, in silico methods are available to investigate protein-protein interactions. Cross-docking methods, in particular, can be used to predict protein binding sites. However, proteins can interact with numerous partners and can present multiple binding sites on their surface, which may alter the binding site prediction quality. We evaluate the binding site predictions obtained using complete cross-docking simulations of 358 proteins with 2 different scoring schemes accounting for multiple binding sites. Despite overall good binding site prediction performances, 68 cases were still associated with very low prediction quality, presenting individual area under the specificity-sensitivity ROC curve (AUC) values below the random AUC threshold of 0.5, since cross-docking calculations can lead to the identification of alternate protein binding sites (that are different from the reference experimental sites). For the large majority of these proteins, we show that the predicted alternate binding sites correspond to interaction sites with hidden partners, that is, partners not included in the original cross-docking dataset. Among those new partners, we find proteins, but also nucleic acid molecules. Finally, for proteins with multiple binding sites on their surface, we investigated the structural determinants associated with the binding sites the most targeted by the docking partners.

Meet-U: Educating through research immersion.

We present a new educational initiative called Meet-U that aims to train students for collaborative work in computational biology and to bridge the gap between education and research. Meet-U mimics the setup of collaborative research projects and takes advantage of the most popular tools for collaborative work and of cloud computing. Students are grouped in teams of 4-5 people and have to realize a project from A to Z that answers a challenging question in biology. Meet-U promotes "coopetition," as the students collaborate within and across the teams and are also in competition with each other to develop the best final product. Meet-U fosters interactions between different actors of education and research through the organization of a meeting day, open to everyone, where the students present their work to a jury of researchers and jury members give research seminars. This very unique combination of education and research is strongly motivating for the students and provides a formidable opportunity for a scientific community to unite and increase its visibility. We report on our experience with Meet-U in two French universities with master's students in bioinformatics and modeling, with protein-protein docking as the subject of the course. Meet-U is easy to implement and can be straightforwardly transferred to other fields and/or universities. All the information and data are available at www.meet-u.org.

Determinants of neuroglobin plasticity highlighted by joint coarse-grained simulations and high pressure crystallography.

Investigating the effect of pressure sheds light on the dynamics and plasticity of proteins, intrinsically correlated to functional efficiency. Here we detail the structural response to pressure of neuroglobin (Ngb), a hexacoordinate globin likely to be involved in neuroprotection. In murine Ngb, reversible coordination is achieved by repositioning the heme more deeply into a large internal cavity, the "heme sliding mechanism". Combining high pressure crystallography and coarse-grain simulations on wild type Ngb as well as two mutants, one (V101F) with unaffected and another (F106W) with decreased affinity for CO, we show that Ngb hinges around a rigid mechanical nucleus of five hydrophobic residues (V68, I72, V109, L113, Y137) during its conformational transition induced by gaseous ligand, that the intrinsic flexibility of the F-G loop appears essential to drive the heme sliding mechanism, and that residue Val 101 may act as a sensor of the interaction disruption between the heme and the distal histidine.

Mobility and Core-Protein Binding Patterns of Disordered C-Terminal Tails in ß-Tubulin Isotypes.

Although they play a significant part in the regulation of microtubule structure, dynamics, and function, the disordered C-terminal tails of tubulin remain invisible to experimental structural methods and do not appear in the crystallographic structures that are currently available in the Protein Data Bank. Interestingly, these tails concentrate most of the sequence variability between tubulin isotypes and are the sites of the principal post-translational modifications undergone by this protein. Using homology modeling, we developed two complete models for the human αI/ßI- and αI/ßIII-tubulin isotypes that include their C-terminal tails. We then investigated the conformational variability of the two ß-tails using long time-scale classical molecular dynamics simulations that revealed similar features, notably the unexpected presence of common anchoring regions on the surface of the tuulin dimer, but also distinctive mobility or interaction patterns, some of which could be related to the tail lengths and charge distributions. We also observed in our simulations that the C-terminal tail from the ßI isotype, but not the ßIII isotype, formed contacts in the putative binding site of a recently discovered peptide that disrupts microtubule formation in glioma cells. Hindering the binding site in the ßI isotype would be consistent with this peptide's preferential disruption of microtubule formation in glioma, whose cells overexpress ßIII, compared to normal glial cells. While these observations need to be confirmed with more intensive sampling, our study opens new perspectives for the development of isotype-specific chemotherapy drugs.

Great interactions: How binding incorrect partners can teach us about protein recognition and function.

Protein-protein interactions play a key part in most biological processes and understanding their mechanism is a fundamental problem leading to numerous practical applications. The prediction of protein binding sites in particular is of paramount importance since proteins now represent a major class of therapeutic targets. Amongst others methods, docking simulations between two proteins known to interact can be a useful tool for the prediction of likely binding patches on a protein surface. From the analysis of the protein interfaces generated by a massive cross-docking experiment using the 168 proteins of the Docking Benchmark 2.0, where all possible protein pairs, and not only experimental ones, have been docked together, we show that it is also possible to predict a protein's binding residues without having any prior knowledge regarding its potential interaction partners. Evaluating the performance of cross-docking predictions using the area under the specificity-sensitivity ROC curve (AUC) leads to an AUC value of 0.77 for the complete benchmark (compared to the 0.5 AUC value obtained for random predictions). Furthermore, a new clustering analysis performed on the binding patches that are scattered on the protein surface show that their distribution and growth will depend on the protein's functional group. Finally, in several cases, the binding-site predictions resulting from the cross-docking simulations will lead to the identification of an alternate interface, which corresponds to the interaction with a biomolecular partner that is not included in the original benchmark. Proteins 2016; 84:1408-1421. © 2016 The Authors Proteins: Structure, Function, and Bioinformatics Published by Wiley Periodicals, Inc.