Representation Of Protein Structure Research Articles

Utilizing Molecular Dynamics Simulations, Machine Learning, Cryo-EM, and NMR Spectroscopy to Predict and Validate Protein Dynamics.

Protein dynamics play a crucial role in biological function, encompassing motions ranging from atomic vibrations to large-scale conformational changes. Recent advancements in experimental techniques, computational methods, and artificial intelligence have revolutionized our understanding of protein dynamics. Nuclear magnetic resonance spectroscopy provides atomic-resolution insights, while molecular dynamics simulations offer detailed trajectories of protein motions. Computational methods applied to X-ray crystallography and cryo-electron microscopy (cryo-EM) have enabled the exploration of protein dynamics, capturing conformational ensembles that were previously unattainable. The integration of machine learning, exemplified by AlphaFold2, has accelerated structure prediction and dynamics analysis. These approaches have revealed the importance of protein dynamics in allosteric regulation, enzyme catalysis, and intrinsically disordered proteins. The shift towards ensemble representations of protein structures and the application of single-molecule techniques have further enhanced our ability to capture the dynamic nature of proteins. Understanding protein dynamics is essential for elucidating biological mechanisms, designing drugs, and developing novel biocatalysts, marking a significant paradigm shift in structural biology and drug discovery.

PUResNetV2.0: a deep learning model leveraging sparse representation for improved ligand binding site prediction

Accurate ligand binding site prediction (LBSP) within proteins is essential for drug discovery. We developed ProteinUNetResNetV2.0 (PUResNetV2.0), leveraging sparse representation of protein structures to improve LBSP accuracy. Our training dataset included protein complexes from 4729 protein families. Evaluations on benchmark datasets showed that PUResNetV2.0 achieved an 85.4% Distance Center Atom (DCA) success rate and a 74.7% F1 Score on the Holo801 dataset, outperforming existing methods. However, its performance in specific cases, such as RNA, DNA, peptide-like ligand, and ion binding site prediction, was limited due to constraints in our training data. Our findings underscore the potential of sparse representation in LBSP, especially for oligomeric structures, suggesting PUResNetV2.0 as a promising tool for computational drug discovery.

Improving AlphaFold Predicted Contacts for Alpha-Helical Transmembrane Proteins Using Structural Features.

Residue contact maps provide a condensed two-dimensional representation of three-dimensional protein structures, serving as a foundational framework in structural modeling but also as an effective tool in their own right in identifying inter-helical binding sites and drawing insights about protein function. Treating contact maps primarily as an intermediate step for 3D structure prediction, contact prediction methods have limited themselves exclusively to sequential features. Now that AlphaFold2 predicts 3D structures with good accuracy in general, we examine (1) how well predicted 3D structures can be directly used for deciding residue contacts, and (2) whether features from 3D structures can be leveraged to further improve residue contact prediction. With a well-known benchmark dataset, we tested predicting inter-helical residue contact based on AlphaFold2's predicted structures, which gave an 83% average precision, already outperforming a sequential features-based state-of-the-art model. We then developed a procedure to extract features from atomic structure in the neighborhood of a residue pair, hypothesizing that these features will be useful in determining if the residue pair is in contact, provided the structure is decently accurate, such as predicted by AlphaFold2. Training on features generated from experimentally determined structures, we leveraged knowledge from known structures to significantly improve residue contact prediction, when testing using the same set of features but derived using AlphaFold2 structures. Our results demonstrate a remarkable improvement over AlphaFold2, achieving over 91.9% average precision for a held-out subset and over 89.5% average precision in cross-validation experiments.

Graphormer supervised de novo protein design method and function validation.

Protein design is central to nearly all protein engineering problems, as it can enable the creation of proteins with new biological functions, such as improving the catalytic efficiency of enzymes. One key facet of protein design, fixed-backbone protein sequence design, seeks to design new sequences that will conform to a prescribed protein backbone structure. Nonetheless, existing sequence design methods present limitations, such as low sequence diversity and shortcomings in experimental validation of the designed functional proteins. These inadequacies obstruct the goal of functional protein design. To improve these limitations, we initially developed the Graphormer-based Protein Design (GPD) model. This model utilizes the Transformer on a graph-based representation of three-dimensional protein structures and incorporates Gaussian noise and a sequence random masks to node features, thereby enhancing sequence recovery and diversity. The performance of the GPD model was significantly better than that of the state-of-the-art ProteinMPNN model on multiple independent tests, especially for sequence diversity. We employed GPD to design CalB hydrolase and generated nine artificially designed CalB proteins. The results show a 1.7-fold increase in catalytic activity compared to that of the wild-type CalB and strong substrate selectivity on p-nitrophenyl acetate with different carbon chain lengths (C2-C16). Thus, the GPD method could be used for the de novo design of industrial enzymes and protein drugs. The code was released at https://github.com/decodermu/GPD.

Q-BioLiP: A Comprehensive Resource for Quaternary Structure-based Protein-ligand Interactions.

Since its establishment in 2013, BioLiP has become one of the widely used resources for protein-ligand interactions. Nevertheless, several known issues occurred with it over the past decade. For example, the protein-ligand interactions are represented in the form of single chain-based tertiary structures, which may be inappropriate as many interactions involve multiple protein chains (known as quaternary structures). We sought to address these issues, resulting in Q-BioLiP, a comprehensive resource for quaternary structure-based protein-ligand interactions. The major features of Q-BioLiP include: (1) representing protein structures in the form of quaternary structures rather than single chain-based tertiary structures; (2) pairing DNA/RNA chains properly rather than separation; (3) providing both experimental and predicted binding affinities; (4) retaining both biologically relevant and irrelevant interactions to alleviate the wrong justification of ligands' biological relevance; and (5) developing a new quaternary structure-based algorithm for the modelling of protein-ligand complex structure. With these new features, Q-BioLiP is expected to be a valuable resource for studying biomolecule interactions, including protein-small molecule interaction, protein-metal ion interaction, protein-peptide interaction, protein-protein interaction, protein-DNA/RNA interaction, and RNA-small molecule interaction. Q-BioLiP is freely available at https://yanglab.qd.sdu.edu.cn/Q-BioLiP/.

SASA-Net: A Spatial-Aware Self-Attention Mechanism for Building Protein 3D Structure Directly From Inter- Residue Distances.

Protein functions are tightly related to the fine details of their 3D structures. To understand protein structures, computational prediction approaches are highly needed. Recently, protein structure prediction has achieved considerable progresses mainly due to the increased accuracy of inter-residue distance estimation and the application of deep learning techniques. Most of the distance-based ab initio prediction approaches adopt a two-step diagram: constructing a potential function based on the estimated inter-residue distances, and then build a 3D structure that minimizes the potential function. These approaches have proven very promising; however, they still suffer from several limitations, especially the inaccuracies incurred by the handcrafted potential function. Here, we present SASA-Net, a deep learning-based approach that directly learns protein 3D structure from the estimated inter-residue distances. Unlike the existing approach simply representing protein structures as coordinates of atoms, SASA-Net represents protein structures using pose of residues, i.e., the coordinate system of each individual residue in which all backbone atoms of this residue are fixed. The key element of SASA-Net is a spatial-aware self-attention mechanism, which is able to adjust a residue's pose according to all other residues' features and the estimated distances between residues. By iteratively applying the spatial-aware self-attention mechanism, SASA-Net continuously improves the structure and finally acquires a structure with high accuracy. Using the CATH35 proteins as representatives, we demonstrate that SASA-Net is able to accurately and efficiently build structures from the estimated inter-residue distances. The high accuracy and efficiency of SASA-Net enables an end-to-end neural network model for protein structure prediction through combining SASA-Net and an neural network for inter-residue distance prediction. Source code of SASA-Net is available at https://github.com/gongtiansu/SASA-Net/.

Two sequence- and two structure-based ML models have learned different aspects of protein biochemistry

Deep learning models are seeing increased use as methods to predict mutational effects or allowed mutations in proteins. The models commonly used for these purposes include large language models (LLMs) and 3D Convolutional Neural Networks (CNNs). These two model types have very different architectures and are commonly trained on different representations of proteins. LLMs make use of the transformer architecture and are trained purely on protein sequences whereas 3D CNNs are trained on voxelized representations of local protein structure. While comparable overall prediction accuracies have been reported for both types of models, it is not known to what extent these models make comparable specific predictions and/or generalize protein biochemistry in similar ways. Here, we perform a systematic comparison of two LLMs and two structure-based models (CNNs) and show that the different model types have distinct strengths and weaknesses. The overall prediction accuracies are largely uncorrelated between the sequence- and structure-based models. Overall, the two structure-based models are better at predicting buried aliphatic and hydrophobic residues whereas the two LLMs are better at predicting solvent-exposed polar and charged amino acids. Finally, we find that a combined model that takes the individual model predictions as input can leverage these individual model strengths and results in significantly improved overall prediction accuracy.

Score-based generative modeling for de novo protein design.

The generation of de novo protein structures with predefined functions and properties remains a challenging problem in protein design. Diffusion models, also known as score-based generative models (SGMs), have recently exhibited astounding empirical performance in image synthesis. Here we use image-based representations of protein structure to develop ProteinSGM, a score-based generative model that produces realistic de novo proteins. Through unconditional generation, we show that ProteinSGM can generate native-like protein structures, surpassing the performance of previously reported generative models. We experimentally validate some de novo designs and observe secondary structure compositions consistent with generated backbones. Finally, we apply conditional generation to de novo protein design by formulating it as an image inpainting problem, allowing precise and modular design of protein structure.

Investigation of Cellular Response to the HSP90 Inhibition in Human Cells Through Thermal Proteome Profiling

Heat shock proteins are chaperones, and they are responsible for protein folding in cells. Heat shock protein 90 (HSP90) is one of the most important chaperones in human cells, and its inhibition is promising for cancer therapy. However, despite the development of multiple HSP90 inhibitors, none of them has been approved for disease treatment due to unexpected cellular toxicity and side effects. Hence, a more comprehensive investigation of cellular response to HSP90 inhibitors can aid in a better understanding of the molecular mechanisms of the cytotoxicity and side effects of these inhibitors. The thermal stability shifts of proteins, which represent protein structure and interaction alterations, can provide valuable information complementary to the results obtained from commonly used abundance-based proteomics analysis. Here, we systematically investigated cell response to different HSP90 inhibitors through global quantification of protein thermal stability changes using thermal proteome profiling, together with the measurement of protein abundance changes. Besides the targets and potential off-targets of the drugs, proteins with significant thermal stability changes under the HSP90 inhibition are found to be involved in cell stress responses and the translation process. Moreover, proteins with thermal stability shifts under the inhibition are upstream of those with altered expression. These findings indicate that the HSP90 inhibition perturbs cell transcription and translation processes. The current study provides a different perspective for achieving a better understanding of cellular response to chaperone inhibition.

Enzyme Substrate Prediction from Three-Dimensional Feature Representations Using Space-Filling Curves.

Compact and interpretable structural feature representations are required for accurately predicting properties and function of proteins. In this work, we construct and evaluate three-dimensional feature representations of protein structures based on space-filling curves (SFCs). We focus on the problem of enzyme substrate prediction, using two ubiquitous enzyme families as case studies: the short-chain dehydrogenase/reductases (SDRs) and the S-adenosylmethionine-dependent methyltransferases (SAM-MTases). Space-filling curves such as the Hilbert curve and the Morton curve generate a reversible mapping from discretized three-dimensional to one-dimensional representations and thus help to encode three-dimensional molecular structures in a system-independent way and with only a few adjustable parameters. Using three-dimensional structures of SDRs and SAM-MTases generated using AlphaFold2, we assess the performance of the SFC-based feature representations in predictions on a new benchmark database of enzyme classification tasks including their cofactor and substrate selectivity. Gradient-boosted tree classifiers yield binary prediction accuracy of 0.77-0.91 and area under curve (AUC) characteristics of 0.83-0.92 for the classification tasks. We investigate the effects of amino acid encoding, spatial orientation, and (the few) parameters of SFC-based encodings on the accuracy of the predictions. Our results suggest that geometry-based approaches such as SFCs are promising for generating protein structural representations and are complementary to the existing protein feature representations such as evolutionary scale modeling (ESM) sequence embeddings.

Integrating Molecular Models Into CryoEM Heterogeneity Analysis Using Scalable High-resolution Deep Gaussian Mixture Models

Resolving the structural variability of proteins is often key to understanding the structure–function relationship of those macromolecular machines. Single particle analysis using Cryogenic electron microscopy (CryoEM), combined with machine learning algorithms, provides a way to reveal the dynamics within the protein system from noisy micrographs. Here, we introduce an improved computational method that uses Gaussian mixture models for protein structure representation and deep neural networks for conformation space embedding. By integrating information from molecular models into the heterogeneity analysis, we can analyze continuous protein conformational changes using structural information at the frequency of 1/3 Å−1, and present the results in a more interpretable form.

GraphSite: Ligand Binding Site Classification with Deep Graph Learning.

The binding of small organic molecules to protein targets is fundamental to a wide array of cellular functions. It is also routinely exploited to develop new therapeutic strategies against a variety of diseases. On that account, the ability to effectively detect and classify ligand binding sites in proteins is of paramount importance to modern structure-based drug discovery. These complex and non-trivial tasks require sophisticated algorithms from the field of artificial intelligence to achieve a high prediction accuracy. In this communication, we describe GraphSite, a deep learning-based method utilizing a graph representation of local protein structures and a state-of-the-art graph neural network to classify ligand binding sites. Using neural weighted message passing layers to effectively capture the structural, physicochemical, and evolutionary characteristics of binding pockets mitigates model overfitting and improves the classification accuracy. Indeed, comprehensive cross-validation benchmarks against a large dataset of binding pockets belonging to 14 diverse functional classes demonstrate that GraphSite yields the class-weighted F1-score of 81.7%, outperforming other approaches such as molecular docking and binding site matching. Further, it also generalizes well to unseen data with the F1-score of 70.7%, which is the expected performance in real-world applications. We also discuss new directions to improve and extend GraphSite in the future.

Fast protein structure comparison through effective representation learning with contrastive graph neural networks.

Protein structure alignment algorithms are often time-consuming, resulting in challenges for large-scale protein structure similarity-based retrieval. There is an urgent need for more efficient structure comparison approaches as the number of protein structures increases rapidly. In this paper, we propose an effective graph-based protein structure representation learning method, GraSR, for fast and accurate structure comparison. In GraSR, a graph is constructed based on the intra-residue distance derived from the tertiary structure. Then, deep graph neural networks (GNNs) with a short-cut connection learn graph representations of the tertiary structures under a contrastive learning framework. To further improve GraSR, a novel dynamic training data partition strategy and length-scaling cosine distance are introduced. We objectively evaluate our method GraSR on SCOPe v2.07 and a new released independent test set from PDB database with a designed comprehensive performance metric. Compared with other state-of-the-art methods, GraSR achieves about 7%-10% improvement on two benchmark datasets. GraSR is also much faster than alignment-based methods. We dig into the model and observe that the superiority of GraSR is mainly brought by the learned discriminative residue-level and global descriptors. The web-server and source code of GraSR are freely available at www.csbio.sjtu.edu.cn/bioinf/GraSR/ for academic use.

Comparison of Benchmark HP Model Proteins on Face-Centered Cubic and Simple Cubic Lattices using Replica-Exchange Wang-Landau Sampling

The HP model represents protein sequences by coarse-graining amino acid types as either hydrophobic (H) or polar (P). This model has been studied extensively, where the simple cubic (sc) lattice is a commonly used for simplicity and efficiency, despite a limited ability to geometrically represent protein structures. One straightforward improvement is to instead use the face-centered cubic (fcc) lattice, which offers a more realistic geometry while retaining simplicity. Using the replica-exchange Wang-Landau algorithm, the ground state energies and full density of states are predicted for a set of biologically-motivated HP sequences on the fcc lattice. A direct thermodynamic comparison is made between the folding on fcc and sc lattices, and structural properties are calculated to further examine the folding on the fcc lattice.

Using Open-Source Bioinformatics and Visualization Tools to Explore the Structure and Function of SARS-CoV-2 Spike Protein

The relationship between protein structure and function is a foundational concept in undergraduate biochemistry. We find this theme is best presented with assignments that encourage exploration and analysis. Here, we share a series of four assignments that use open-source, online molecular visualization and bioinformatics tools to examine the interaction between the SARS-CoV-2 spike protein and the ACE2 receptor. The interaction between these two proteins initiates SARS-CoV-2 infection of human host cells and is the cause of COVID-19. In assignment I, students identify sequences with homology to the SARS-CoV-2 spike protein and use them to build a primary sequence alignment. Students make connections to a linked primary research article as an example of how scientists use molecular and phylogenetic analysis to explore the origins of a novel virus. Assignments II through IV teach students to use an online molecular visualization tool for analysis of secondary, tertiary, and quaternary structure. Emphasis is placed on identification of noncovalent interactions that stabilize the SARS-CoV-2 spike protein and mediate its interaction with ACE2. We assigned this project to upper-level undergraduate biochemistry students at a public university and liberal arts college. Students in our courses completed the project as individual homework assignments. However, we can easily envision implementation of this project during multiple in-class sessions or in a biochemistry laboratory using in-person or remote learning. We share this project as a resource for instructors who aim to teach protein structure and function using inquiry-based molecular visualization activities. <em>Primary image:</em>&nbsp;Exploration of SARS-CoV-2 spike protein: student generated data from assignments I - IV. Includes examples of figures submitted by students, including a sequence alignment and representations of 3D protein structure generated using UCSF Chimera.&nbsp;The primary image includes student generated data and a cartoon from Pixabay, an online repository of copyright free art.&nbsp;

Prediction of GPCR activity using machine learning

GPCRs are the target for one-third of the FDA-approved drugs, however; the development of new drug molecules targeting GPCRs is limited by the lack of mechanistic understanding of the GPCR structure–activity-function relationship. To modulate the GPCR activity with highly specific drugs and minimal side-effects, it is necessary to quantitatively describe the important structural features in the GPCR and correlate them to the activation state of GPCR. In this study, we developed 3 ML approaches to predict the conformation state of GPCR proteins. Additionally, we predict the activity level of GPCRs based on their structure. We leverage the unique advantages of each of the 3 ML approaches, interpretability of XGBoost, minimal feature engineering for 3D convolutional neural network, and graph representation of protein structure for graph neural network. By using these ML approaches, we are able to predict the activation state of GPCRs with high accuracy (91%–95%) and also predict the activation state of GPCRs with low error (MAE of 7.15–10.58). Furthermore, the interpretation of the ML approaches allows us to determine the importance of each of the features in distinguishing between the GPCRs conformations.

Cosolvent and Dynamic Effects in Binding Pocket Search by Docking Simulations.

The lack of conformational sampling in virtual screening projects can lead to inefficient results because many of the potential drugs may not be able to bind to the target protein during the static docking simulations. Here, we performed ensemble docking for around 2000 United States Food and Drug Administration (FDA)-approved drugs with the RNA-dependent RNA polymerase (RdRp) protein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) as a target. The representative protein structures were generated by clustering classical molecular dynamics trajectories, which were evolved using three solvent scenarios, namely, pure water, benzene/water and phenol/water mixtures. The introduction of dynamic effects in the theoretical model showed improvement in docking results in terms of the number of strong binders and binding sites in the protein. Some of the discovered pockets were found only for the cosolvent simulations, where the nonpolar probes induced local conformational changes in the protein that lead to the opening of transient pockets. In addition, the selection of the ligands based on a combination of the binding free energy and binding free energy gap between the best two poses for each ligand provided more suitable binders than the selection of ligands based solely on one of the criteria. The application of cosolvent molecular dynamics to enhance the sampling of the configurational space is expected to improve the efficacy of virtual screening campaigns of future drug discovery projects.

XENet: Using a new graph convolution to accelerate the timeline for protein design on quantum computers.

Graph representations are traditionally used to represent protein structures in sequence design protocols in which the protein backbone conformation is known. This infrequently extends to machine learning projects: existing graph convolution algorithms have shortcomings when representing protein environments. One reason for this is the lack of emphasis on edge attributes during massage-passing operations. Another reason is the traditionally shallow nature of graph neural network architectures. Here we introduce an improved message-passing operation that is better equipped to model local kinematics problems such as protein design. Our approach, XENet, pays special attention to both incoming and outgoing edge attributes. We compare XENet against existing graph convolutions in an attempt to decrease rotamer sample counts in Rosetta's rotamer substitution protocol, used for protein side-chain optimization and sequence design. This use case is motivating because it both reduces the size of the search space for classical side-chain optimization algorithms, and allows larger protein design problems to be solved with quantum algorithms on near-term quantum computers with limited qubit counts. XENet outperformed competing models while also displaying a greater tolerance for deeper architectures. We found that XENet was able to decrease rotamer counts by 40% without loss in quality. This decreased the memory consumption for classical pre-computation of rotamer energies in our use case by more than a factor of 3, the qubit consumption for an existing sequence design quantum algorithm by 40%, and the size of the solution space by a factor of 165. Additionally, XENet displayed an ability to handle deeper architectures than competing convolutions.

Protein-Peptide Binding Site Detection Using 3D Convolutional Neural Networks.

Peptides and peptide-based molecules represent a promising therapeutic modality targeting intracellular protein-protein interactions, potentially combining the beneficial properties of biologics and small-molecule drugs. Protein-peptide complexes occupy a unique niche of interaction interfaces with respect to protein-protein and protein-small molecule complexes. Protein-peptide binding site identification resembles image object detection, a field that had been revolutionalized with computer vision techniques. We present a new protein-peptide binding site detection method called BiteNetPp by harnessing the power of 3D convolutional neural network. Our method employs a tensor-based representation of spatial protein structures, which is fed to 3D convolutional neural network, resulting in probability scores and coordinates of the binding "hot spots" in the input structures. We used the domain adaptation technique to fine-tune model trained on protein-small molecule complexes using a manually curated set of protein-peptide structures. BiteNetPp consistently outperforms existing state-of-the-art methods in the independent test benchmark. It takes less than a second to analyze a single-protein structure, making BiteNetPp suitable for the large-scale analysis of protein-peptide binding sites.

SidechainNet: An all‐atom protein structure dataset for machine learning

Despite recent advancements in deep learning methods for protein structure prediction and representation, little focus has been directed at the simultaneous inclusion and prediction of protein backbone and sidechain structure information. We present SidechainNet, a new dataset that directly extends the ProteinNet dataset. SidechainNet includes angle and atomic coordinate information capable of describing all heavy atoms of each protein structure and can be extended by users to include new protein structures as they are released. In this article, we provide background information on the availability of protein structure data and the significance of ProteinNet. Thereafter, we argue for the potentially beneficial inclusion of sidechain information through SidechainNet, describe the process by which we organize SidechainNet, and provide a software package (https://github.com/jonathanking/sidechainnet) for data manipulation and training with machine learning models.