Protein loop structure prediction by community-based deep learning and its application to antibody CDR H3 loop modeling.

As of now, more than 60 years have passed since the first determination of protein structures through crystallography, and a significant portion of protein structures can be predicted by computers. This is due to the groundbreaking enhancement in protein structure prediction achieved through neural network training utilizing extensive sequence and structure data. However, substantial challenges persist in structure prediction due to limited data availability, with antibody structure prediction standing as one such challenge. In this paper, we propose a novel neural network architecture that effectively enables structure prediction by reflecting the inherent combinatorial nature involved in protein structure formation. The core idea of this neural network architecture is not solely to track and generate a single structure but rather to form a community of multiple structures and pursue accurate structure prediction by exchanging information among community members. Applying this concept to antibody CDR H3 loop structure prediction resulted in improved structure sampling. Such an approach could be applied in the structural and functional studies of proteins, particularly in exploring various physiological processes mediated by loops. Moreover, it holds potential in addressing various other types of combinatorial structure prediction and design problems.

HProteome-BSite: predicted binding sites and ligands in human 3D proteome.

Atomic-level knowledge of protein-ligand interactions allows a detailed understanding of protein functions and provides critical clues to discovering molecules regulating the functions. While recent innovative deep learning methods for protein structure prediction dramatically increased the structural coverage of the human proteome, molecular interactions remain largely unknown. A new database, HProteome-BSite, provides predictions of binding sites and ligands in the enlarged 3D human proteome. The model structures for human proteins from the AlphaFold Protein Structure Database were processed to structural domains of high confidence to maximize the coverage and reliability of interaction prediction. For ligand binding site prediction, an updated version of a template-based method GalaxySite was used. A high-level performance of the updated GalaxySite was confirmed. HProteome-BSite covers 80.74% of the UniProt entries in the AlphaFold human 3D proteome. Predicted binding sites and binding poses of potential ligands are provided for effective applications to further functional studies and drug discovery. The HProteome-BSite database is available at https://galaxy.seoklab.org/hproteome-bsite/database and is free and open to all users.

Ebastine impairs metastatic spread in triple-negative breast cancer by targeting focal adhesion kinase.

We sought to investigate the utility of ebastine (EBA), a second-generation antihistamine with potent anti-metastatic properties, in the context of breast cancer stem cell (BCSC)-suppression in triple-negative breast cancer (TNBC). EBA binds to the tyrosine kinase domain of focal adhesion kinase (FAK), blocking phosphorylation at the Y397 and Y576/577 residues. FAK-mediated JAK2/STAT3 and MEK/ERK signaling was attenuated after EBA challenge in vitro and in vivo. EBA treatment induced apoptosis and a sharp decline in the expression of the BCSC markers ALDH1, CD44 and CD49f, suggesting that EBA targets BCSC-like cell populations while reducing tumor bulk. EBA administration significantly impeded BCSC-enriched tumor burden, angiogenesis and distant metastasis while reducing MMP-2/-9 levels in circulating blood in vivo. Our findings suggest that EBA may represent an effective therapeutic for the simultaneous targeting of JAK2/STAT3 and MEK/ERK for the treatment of molecularly heterogeneous TNBC with divergent profiles. Further investigation of EBA as an anti-metastatic agent for the treatment of TNBC is warranted.

A pGpG-specific phosphodiesterase regulates cyclic di-GMP signaling in Vibrio cholerae.

The bacterial second messenger bis-(3'-5')-cyclic diguanylate monophosphate (c-di-GMP) controls various cellular processes, including motility, toxin production, and biofilm formation. c-di-GMP is enzymatically synthesized by GGDEF domain-containing diguanylate cyclases and degraded by HD-GYP domain-containing phosphodiesterases (PDEs) to 2 GMP or by EAL domain-containing PDE-As to 5'-phosphoguanylyl-(3',5')-guanosine (pGpG). Since excess pGpG feedback inhibits PDE-A activity and thereby can lead to the uncontrolled accumulation of c-di-GMP, a PDE that degrades pGpG to 2 GMP (PDE-B) has been presumed to exist. To date, the only enzyme known to hydrolyze pGpG is oligoribonuclease Orn, which degrades all kinds of oligoribonucleotides. Here, we identified a pGpG-specific PDE, which we named PggH, using biochemical approaches in the gram-negative bacteria Vibrio cholerae. Biochemical experiments revealed that PggH exhibited specific PDE activity only toward pGpG, thus differing from the previously reported Orn. Furthermore, the high-resolution structure of PggH revealed the basis for its PDE activity and narrow substrate specificity. Finally, we propose that PggH could modulate the activities of PDE-As and the intracellular concentration of c-di-GMP, resulting in phenotypic changes including in biofilm formation.

Atomic-level thermodynamics analysis of the binding free energy of SARS-CoV-2 neutralizing antibodies.

Understanding how protein-protein binding affinity is determined from molecular interactions at the interface is essential in developing protein therapeutics such as antibodies, but this has not yet been fully achieved. Among the major difficulties are the facts that it is generally difficult to decompose thermodynamic quantities into contributions from individual molecular interactions and that the solvent effect-dehydration penalty-must also be taken into consideration for every contact formation at the binding interface. Here, we present an atomic-level thermodynamics analysis that overcomes these difficulties and illustrate its utility through application to SARS-CoV-2 neutralizing antibodies. Our analysis is based on the direct interaction energy computed from simulated antibody-protein complex structures and on the decomposition of solvation free energy change upon complex formation. We find that the formation of a single contact such as a hydrogen bond at the interface barely contributes to binding free energy due to the dehydration penalty. On the other hand, the simultaneous formation of multiple contacts between two interface residues favorably contributes to binding affinity. This is because the dehydration penalty is significantly alleviated: the total penalty for multiple contacts is smaller than a sum of what would be expected for individual dehydrations of those contacts. Our results thus provide a new perspective for designing protein therapeutics of improved binding affinity.

Benchmarking applicability of medium-resolution cryo-EM protein structures for structure-based drug design.

Cryo-electron microscopy (cryo-EM) is gaining large attention for high-resolution protein structure determination in solutions. However, a very high percentage of cryo-EM structures correspond to resolutions of 3-5 Å, making the structures difficult to be used in in silico drug design. In this study, we analyze how useful cryo-EM protein structures are for in silico drug design by evaluating ligand docking accuracy. From realistic cross-docking scenarios using medium resolution (3-5 Å) cryo-EM structures and a popular docking tool Autodock-Vina, only 20% of docking succeeded, when the success rate doubles in the same kind of cross-docking but using high-resolution (<2 Å) crystal structures instead. We decipher the reason for failures by decomposing the contribution from resolution-dependent and independent factors. The heterogeneity in the protein side-chain and backbone conformations is identified as the major resolution-dependent factor causing docking difficulty from our analysis, while intrinsic receptor flexibility mainly comprises the resolution-independent factor. We demonstrate the flexibility implementation in current ligand docking tools is able to rescue only a portion of failures (10%), and the limited performance was majorly due to potential structural errors than conformational changes. Our work suggests the strong necessity of more robust method developments on ligand docking and EM modeling techniques in order to fully utilize cryo-EM structures for in silico drug design.

GalaxyDock2-HEME: Protein-ligand docking for heme proteins.

Prediction of protein-ligand binding poses is an essential component for understanding protein-ligand interactions and computer-aided drug design. Various proteins involve prosthetic groups such as heme for their functions, and adequate consideration of the prosthetic groups is vital for protein-ligand docking. Here, we extend the GalaxyDock2 protein-ligand docking algorithm to handle ligand docking to heme proteins. Docking to heme proteins involves increased complexity because the interaction of heme iron and ligand has covalent nature. GalaxyDock2-HEME, a new protein-ligand docking program for heme proteins, has been developed based on GalaxyDock2 by adding an orientation-dependent scoring term to describe heme iron-ligand coordination interaction. This new docking program performs better than other noncommercial docking programs such as EADock with MMBP, AutoDock Vina, PLANTS, LeDock, and GalaxyDock2 on a heme protein-ligand docking benchmark set in which ligands are known to bind iron. In addition, docking results on two other sets of heme protein-ligand complexes in which ligands do not bind iron show that GalaxyDock2-HEME does not have a high bias toward iron binding compared to other docking programs. This implies that the new docking program can distinguish iron binders from noniron binders for heme proteins.

GalaxyHeteromer: protein heterodimer structure prediction by template-based and ab initio docking.

Protein-protein interactions play crucial roles in diverse biological processes, including various disease progressions. Atomistic structural details of protein-protein interactions may provide important information that can facilitate the design of therapeutic agents. GalaxyHeteromer is a freely available automatic web server (http://galaxy.seoklab.org/heteromer) that predicts protein heterodimer complex structures from two subunit protein sequences or structures. When subunit structures are unavailable, they are predicted by template- or distance-prediction-based modelling methods. Heterodimer complex structures can be predicted by both template-based and ab initio docking, depending on the template's availability. Structural templates are detected from the protein structure database based on both the sequence and structure similarities. The templates for heterodimers may be selected from monomer and homo-oligomer structures, as well as from hetero-oligomers, owing to the evolutionary relationships of heterodimers with domains of monomers or subunits of homo-oligomers. In addition, the server employs one of the best ab initio docking methods when heterodimer templates are unavailable. The multiple heterodimer structure models and the associated scores, which are provided by the web server, may be further examined by user to test or develop functional hypotheses or to design new functional molecules.

GalaxyWater-CNN: Prediction of Water Positions on the Protein Structure by a 3D-Convolutional Neural Network.

Proteins interact with numerous water molecules to perform their physiological functions in biological organisms. Most water molecules act as solvent media; hence, their roles may be considered implicitly in theoretical treatments of protein structure and function. However, some water molecules interact intimately with proteins and require explicit treatment to understand their effects. Most physics-based computational methods are limited in their ability to accurately locate water molecules on protein surfaces because of inaccurate energy functions. Instead of relying on an energy function, this study attempts to learn the locations of water molecules from structural data. GalaxyWater-convolutional neural network (CNN) predicts water positions on protein chains, protein-protein interfaces, and protein-compound binding sites using a 3D-CNN model that is trained to generate a water score map on a given protein structure. The training data are compiled from high-resolution protein crystal structures resolved together with water molecules. GalaxyWater-CNN shows improved water prediction performance both in the coverage of crystal water molecules and in the accuracy of the predicted water positions when compared with previous energy-based methods. This method shows a superior performance in predicting water molecules that form hydrogen-bond networks precisely. The web service and the source code of this water prediction method are freely available at https://galaxy.seoklab.org/gwcnn and https://github.com/seoklab/GalaxyWater-CNN, respectively.

Assessment of protein model structure accuracy estimation in CASP14: Old and new challenges.

In CASP, blind testing of model accuracy estimation methods has been conducted on models submitted by tertiary structure prediction servers. In CASP14, model accuracy estimation results were evaluated in terms of both global and local structure accuracy, as in the previous CASPs. Unlike the previous CASPs that did not show pronounced improvements in performance, the best single-model method (from the Baker group) showed an improved performance in CASP14, particularly in evaluating global structure accuracy when compared to both the best single-model methods in previous CASPs and the best multi-model methods in the current CASP. Although the CASP14 experiment on model accuracy estimation did not deal with the structures generated by AlphaFold2, new challenges that have arisen due to the success of AlphaFold2 are discussed.

Protein oligomer structure prediction using GALAXY in CASP14.

Proteins perform their functions by interacting with other biomolecules. For these interactions, proteins often form homo- or hetero-oligomers as well. Thus, oligomer protein structures provide important clues regarding the biological roles of proteins. To this end, computational prediction of oligomer structures may be a useful tool in the absence of experimentally resolved structures. Here, we describe our server and human-expert methods used to predict oligomer structures in the CASP14 experiment. Examples are provided for cases in which manual domain-splitting led to improved oligomeric domain structures by ab initio docking, automated oligomer structure refinement led to improved subunit orientation and terminal structure, and manual oligomer modeling utilizing literature information generated a reasonable oligomer model. We also discussed the results of post-prediction docking calculations with AlphaFold2 monomers as input in comparison to our blind prediction results. Overall, ab initio docking of AlphaFold2 models did not lead to better oligomer structure prediction, which may be attributed to the interfacial structural difference between the AlphaFold2 monomer structures and the crystal oligomer structures. This result poses a next-stage challenge in oligomer structure prediction after the success of AlphaFold2. For successful protein assembly structure prediction, a different approach that exploits further evolutionary information on the interface and/or flexible docking taking the interfacial conformational flexibilities of subunit structures into account is needed.

Modeling SARS-CoV-2 proteins in the CASP-commons experiment.

Critical Assessment of Structure Prediction (CASP) is an organization aimed at advancing the state of the art in computing protein structure from sequence. In the spring of 2020, CASP launched a community project to compute the structures of the most structurally challenging proteins coded for in the SARS-CoV-2 genome. Forty-seven research groups submitted over 3000 three-dimensional models and 700 sets of accuracy estimates on 10 proteins. The resulting models were released to the public. CASP community members also worked together to provide estimates of local and global accuracy and identify structure-based domain boundaries for some proteins. Subsequently, two of these structures (ORF3a and ORF8) have been solved experimentally, allowing assessment of both model quality and the accuracy estimates. Models from the AlphaFold2 group were found to have good agreement with the experimental structures, with main chain GDT_TS accuracy scores ranging from 63 (a correct topology) to 87 (competitive with experiment).

GalaxyWater-wKGB: Prediction of Water Positions on Protein Structure Using wKGB Statistical Potential.

Proteins fold and function in water, and protein-water interactions play important roles in protein structure and function. In computational studies on protein structure and interaction, the effect of water is considered either implicitly or explicitly. Implicit water models are frequently used in protein structure prediction and docking because they are computationally much more efficient than explicit water models, which are often employed in molecular dynamics (MD) simulations. However, implicit water models that treat water as a continuous solvent medium cannot account for specific atomistic protein-water interactions that are critical for structure formation and interactions with other molecules. Various methods for predicting water molecules that form specific atomistic interactions with proteins have been developed. Methods involving MD simulations or the integral equation theory tend to produce more accurate results at a higher computational cost than simple geometry- or energy-based methods. Here, we present a novel method for predicting water positions on a protein surface called GalaxyWater-wKGB, which is based on a statistical potential, a water knowledge-based potential based on the generalized Born model (wKGB). This method is accurate and rapid because it does not require conformational sampling or iterative computation owing to the effective statistical treatment employed to derive the potential. The statistical potential describes specific protein atom-water interactions more accurately than conventional potentials by considering the dependence on the degree of solvent accessibility of protein atoms as well as on protein atom-water distances and orientations. The introduction of solvent accessibility allows effective consideration of competing nonspecific protein-water and intraprotein interactions. When tested on high-resolution protein crystal structures, this method could recover similar or larger fractions of crystallographic water 180 times faster than the sophisticated integral equation theory, 3D-RISM. A web service of this water prediction method is freely available at http://galaxy.seoklab.org/wkgb.

GalaxyRefine2: simultaneous refinement of inaccurate local regions and overall protein structure.

The 3D structure of a protein can be predicted from its amino acid sequence with high accuracy for a large fraction of cases because of the availability of large quantities of experimental data and the advance of computational algorithms. Recently, deep learning methods exploiting the coevolution information obtained by comparing related protein sequences have been successfully used to generate highly accurate model structures even in the absence of template structure information. However, structures predicted based on either template structures or related sequences require further improvement in regions for which information is missing. Refining a predicted protein structure with insufficient information on certain regions is critical because these regions may be connected to functional specificity that is not conserved among related proteins. The GalaxyRefine2 web server, freely available via http://galaxy.seoklab.org/refine2, is an upgraded version of the GalaxyRefine protein structure refinement server and reflects recent developments successfully tested through CASP blind prediction experiments. This method adopts an iterative optimization approach involving various structure move sets to refine both local and global structures. The estimation of local error and hybridization of available homolog structures are also employed for effective conformation search.

Biophysical and functional characterization of Norrin signaling through Frizzled4.

Wnt signaling is initiated by Wnt ligand binding to the extracellular ligand binding domain, called the cysteine-rich domain (CRD), of a Frizzled (Fzd) receptor. Norrin, an atypical Fzd ligand, specifically interacts with Fzd4 to activate ß-catenin-dependent canonical Wnt signaling. Much of the molecular basis that confers Norrin selectivity in binding to Fzd4 was revealed through the structural study of the Fzd4CRD-Norrin complex. However, how the ligand interaction, seemingly localized at the CRD, is transmitted across full-length Fzd4 to the cytoplasm remains largely unknown. Here, we show that a flexible linker domain, which connects the CRD to the transmembrane domain, plays an important role in Norrin signaling. The linker domain directly contributes to the high-affinity interaction between Fzd4 and Norrin as shown by ∼10-fold higher binding affinity of Fzd4CRD to Norrin in the presence of the linker. Swapping the Fzd4 linker with the Fzd5 linker resulted in the loss of Norrin signaling, suggesting the importance of the linker in ligand-specific cellular response. In addition, structural dynamics of Fzd4 associated with Norrin binding investigated by hydrogen/deuterium exchange MS revealed Norrin-induced conformational changes on the linker domain and the intracellular loop 3 (ICL3) region of Fzd4. Cell-based functional assays showed that linker deletion, L430A and L433A mutations at ICL3, and C-terminal tail truncation displayed reduced ß-catenin-dependent signaling activity, indicating the functional significance of these sites. Together, our results provide functional and biochemical dissection of Fzd4 in Norrin signaling.

Structure prediction of biological assemblies using GALAXY in CAPRI rounds 38-45.

We participated in CARPI rounds 38-45 both as a server predictor and a human predictor. These CAPRI rounds provided excellent opportunities for testing prediction methods for three classes of protein interactions, that is, protein-protein, protein-peptide, and protein-oligosaccharide interactions. Both template-based methods (GalaxyTBM for monomer protein, GalaxyHomomer for homo-oligomer protein, GalaxyPepDock for protein-peptide complex) and ab initio docking methods (GalaxyTongDock and GalaxyPPDock for protein oligomer, GalaxyPepDock-ab-initio for protein-peptide complex, GalaxyDock2 and Galaxy7TM for protein-oligosaccharide complex) have been tested. Template-based methods depend heavily on the availability of proper templates and template-target similarity, and template-target difference is responsible for inaccuracy of template-based models. Inaccurate template-based models could be improved by our structure refinement and loop modeling methods based on physics-based energy optimization (GalaxyRefineComplex and GalaxyLoop) for several CAPRI targets. Current ab initio docking methods require accurate protein structures as input. Small conformational changes from input structure could be accounted for by our docking methods, producing one of the best models for several CAPRI targets. However, predicting large conformational changes involving protein backbone is still challenging, and full exploration of physics-based methods for such problems is still to come.

Prediction of Molecular Electronic Transitions Using Random Forests.

Fluorescent molecules, fluorophores or dyes, play essential roles in bioimaging. Effective bioimaging requires fluorophores with diverse colors and high quantum yields for better resolution. An essential computational component to design novel dye molecules is an accurate model that predicts the electronic properties of molecules. Here, we present statistical machines that predict the excitation energies and associated oscillator strengths of a given molecule using the random forest algorithm. The excitation energies and oscillator strengths of a molecule are closely related to the emission spectrum and the quantum yields of fluorophores, respectively. In this study, we identified specific molecular substructures that induce high oscillator strengths of molecules. The results of our study are expected to serve as new design principles for designing novel fluorophores.

GalaxySagittarius: Structure- and Similarity-Based Prediction of Protein Targets for Druglike Compounds.

Computational techniques for predicting interactions of proteins and druglike molecules have often been used to search for compounds that bind a given protein with high affinity. More recently, such tools have also been applied to the reverse procedure of searching protein targets for a given compound. Among methods for predicting protein-ligand interactions, ligand-based methods relying on similarity to ligands of known interactions are effective only when similar protein-ligand interactions are known. Receptor-based methods predicting protein-ligand interactions by molecular docking are effective only when high-accuracy receptor structures and binding sites are available. Moreover, the computational cost of molecular docking tends to be too high to be applied to the entire protein structure database. In this paper, an effective target prediction method, which combines ligand similarity-based and receptor structure-based approaches, is introduced. In this method, protein-ligand docking is performed after efficient structure- and similarity-based screening. The enriched protein target database by predicted binding ligands and sites allows detection of protein targets with previously unknown ligand interactions. The method, called GalaxySagittarius, is freely available as a web server at http://galaxy.seoklab.org/sagittarius.

Assessment of protein model structure accuracy estimation in CASP13: Challenges in the era of deep learning.

Scoring model structure is an essential component of protein structure prediction that can affect the prediction accuracy tremendously. Users of protein structure prediction results also need to score models to select the best models for their application studies. In Critical Assessment of techniques for protein Structure Prediction (CASP), model accuracy estimation methods have been tested in a blind fashion by providing models submitted by the tertiary structure prediction servers for scoring. In CASP13, model accuracy estimation results were evaluated in terms of both global and local structure accuracy. Global structure accuracy estimation was evaluated by the quality of the models selected by the global structure scores and by the absolute estimates of the global scores. Residue-wise, local structure accuracy estimations were evaluated by three different measures. A new measure introduced in CASP13 evaluates the ability to predict inaccurately modeled regions that may be improved by refinement. An intensive comparative analysis on CASP13 and the previous CASPs revealed that the tertiary structure models generated by the CASP13 servers show very distinct features. Higher consensus toward models of higher global accuracy appeared even for free modeling targets, and many models of high global accuracy were not well optimized at the atomic level. This is related to the new technology in CASP13, deep learning for tertiary contact prediction. The tertiary model structures generated by deep learning pose a new challenge for EMA (estimation of model accuracy) method developers. Model accuracy estimation itself is also an area where deep learning can potentially have an impact, although current EMA methods have not fully explored that direction.

Prediction of protein oligomer structures using GALAXY in CASP13.

Many proteins need to form oligomers to be functional, so oligomer structures provide important clues to biological roles of proteins. Prediction of oligomer structures therefore can be a useful tool in the absence of experimentally resolved structures. In this article, we describe the server and human methods that we used to predict oligomer structures in the CASP13 experiment. Performances of the methods on the 42 CASP13 oligomer targets consisting of 30 homo-oligomers and 12 hetero-oligomers are discussed. Our server method, Seok-assembly, generated models with interface contact similarity measure greater than 0.2 as model 1 for 11 homo-oligomer targets when proper templates existed in the database. Model refinement methods such as loop modeling and molecular dynamics (MD)-based overall refinement failed to improve model qualities when target proteins have domains not covered by templates or when chains have very small interfaces. In human predictions, additional experimental data such as low-resolution electron microscopy (EM) map were utilized. EM data could assist oligomer structure prediction by providing a global shape of the complex structure.