Protein Structure Prediction Methods Research Articles

CHESS 3: an improved, comprehensive catalog of human genes and transcripts based on large-scale expression data, phylogenetic analysis, and protein structure

CHESS 3 represents an improved human gene catalog based on nearly 10,000 RNA-seq experiments across 54 body sites. It significantly improves current genome annotation by integrating the latest reference data and algorithms, machine learning techniques for noise filtering, and new protein structure prediction methods. CHESS 3 contains 41,356 genes, including 19,839 protein-coding genes and 158,377 transcripts, with 14,863 protein-coding transcripts not in other catalogs. It includes all MANE transcripts and at least one transcript for most RefSeq and GENCODE genes. On the CHM13 human genome, the CHESS 3 catalog contains an additional 129 protein-coding genes. CHESS 3 is available at http://ccb.jhu.edu/chess.

Comparing Supervised Learning and Rigorous Approach for Predicting Protein Stability upon Point Mutations in Difficult Targets.

Accurate prediction of protein stability upon a point mutation has important applications in drug discovery and personalized medicine. It remains a challenging issue in computational biology. Existing computational prediction methods, which range from mechanistic to supervised learning approaches, have experienced limited progress over the last few decades. This stagnation is largely due to their heavy reliance on both the quantity and quality of the training data. This is evident in recent state-of-the-art methods that continue to yield substantial errors on two challenging blind test sets: frataxin and p53, with average root-mean-square errors exceeding 3 and 1.5 kcal/mol, respectively, which is still above the theoretical 1 kcal/mol prediction barrier. Rigorous approaches, on the other hand, offer greater potential for accuracy without relying on training data but are computationally demanding and require both wild-type and mutant structure information. Although they showed high accuracy for conserving mutations, their performance is still limited for charge-changing mutation cases. This might be due to the lack of an available mutant structure, often represented by a simplified capped peptide. The recent advances in protein structure prediction methods now make it possible to obtain structures comparable to experimental ones, including complete mutant structure information. In this work, we compare the performance of supervised learning-based methods and rigorous approaches for predicting protein stability on point mutations in difficult targets: frataxin and p53. The rigorous alchemical method significantly surpasses state-of-the-art techniques in terms of both the root-mean-squared error and Pearson correlation coefficient in these two challenging blind test sets. Additionally, we propose an improved alchemical method that employs the pmx double-system/single-box approach to accurately predict the folding free energy change upon both conserving and charge-changing mutations. The enhanced protocol can accurately predict both types of mutations, thereby outperforming existing state-of-the-art methods in overall performance.

Structural and molecular biology of Sabiá virus.

Brazilian mammarenavirus, or Sabiá virus (SABV), is a New World (NW) arenavirus associated with fulminant hemorrhagic disease in humans and the sole biosafety level 4 microorganism ever isolated in Brazil. Since the isolation of SABV in the 1990s, studies on viral biology have been scarce, with no available countermeasures against SABV infection or disease. Here we provide a comprehensive review of SABV biology, including key aspects of SABV replication, and comparisons with related Old World and NW arenaviruses. SABV is most likely a rodent-borne virus, transmitted to humans, through exposure to urine and feces in peri-urban areas. Using protein structure prediction methods and alignments, we analyzed shared and unique features of SABV proteins (GPC, NP, Z, and L) that could be explored in search of therapeutic strategies, including repurposing intended application against arenaviruses. Highly conserved catalytic activities present in L protein could be targeted for broad-acting antiviral activity among arenaviruses, while protein-protein interactions, such as those between L and the matrix protein Z, have evolved in NW arenaviruses and should be specific to SABV. The nucleoprotein (NP) also shares targetable interaction interfaces with L and Z and exhibits exonuclease activity in the C-terminal domain, which may be involved in multiple aspects of SABV replication. Envelope glycoproteins GP1 and GP2 have been explored in the development of promising cross-reactive neutralizing antibodies and vaccines, some of which could be repurposed for SABV. GP1 remains a challenging target in SABV as evolutive pressures render it the most variable viral protein in terms of both sequence and structure, while antiviral strategies targeting the Z protein remain to be validated. In conclusion, the prediction and analysis of protein structures should revolutionize research on viruses such as SABV by facilitating the rational design of countermeasures while reducing dependence on sophisticated laboratory infrastructure for experimental validation.

ComplexQA: a deep graph learning approach for protein complex structure assessment.

In recent years, the end-to-end deep learning method for single-chain protein structure prediction has achieved high accuracy. For example, the state-of-the-art method AlphaFold, developed by Google, has largely increased the accuracy of protein structure predictions to near experimental accuracy in some of the cases. At the same time, there are few methods that can evaluate the quality of protein complexes at the residue level. In particular, evaluating the quality of residues at the interface of protein complexes can lead to a wide range of applications, such as protein function analysis and drug design. In this paper, we introduce a new deep graph neural network-based method ComplexQA, to evaluate the local quality of interfaces for protein complexes by utilizing the residue-level structural information in 3D space and the sequence-level constraints. We benchmark our method to other state-of-the-art quality assessment approaches on the HAF2 and DBM55-AF2 datasets (high-quality structural models predicted by AlphaFold-Multimer), and the BM5 docking dataset. The experimental results show that our proposed method achieves better or similar performance compared with other state-of-the-art methods, especially on difficult targets which only contain a few acceptable models. Our method is able to suggest a score for each interfac e residue, which demonstrates a powerful assessment tool for the ever-increasing number of protein complexes. https://github.com/Cao-Labs/ComplexQA.git. Contact: caora@plu.edu.

Bayesian methods in integrative structure modeling.

There is a growing interest in characterizing the structure and dynamics of large biomolecular assemblies and their interactions within the cellular environment. A diverse array of experimental techniques allows us to study biomolecular systems on a variety of length and time scales. These techniques range from imaging with light, X-rays or electrons, to spectroscopic methods, cross-linking mass spectrometry and functional genomics approaches, and are complemented by AI-assisted protein structure prediction methods. A challenge is to integrate all of these data into a model of the system and its functional dynamics. This review focuses on Bayesian approaches to integrative structure modeling. We sketch the principles of Bayesian inference, highlight recent applications to integrative modeling and conclude with a discussion of current challenges and future perspectives.

XProtCAS: A Toolkit for Extracting Conserved Accessible Surfaces from Protein Structures.

The identification of protein surfaces required for interaction with other biomolecules broadens our understanding of protein function, their regulation by post-translational modification, and the deleterious effect of disease mutations. Protein interaction interfaces are often identifiable as patches of conserved residues on a protein's surface. However, finding conserved accessible surfaces on folded regions requires an understanding of the protein structure to discriminate between functional and structural constraints on residue conservation. With the emergence of deep learning methods for protein structure prediction, high-quality structural models are now available for any protein. In this study, we introduce tools to identify conserved surfaces on AlphaFold2 structural models. We define autonomous structural modules from the structural models and convert these modules to a graph encoding residue topology, accessibility, and conservation. Conserved surfaces are then extracted using a novel eigenvector centrality-based approach. We apply the tool to the human proteome identifying hundreds of uncharacterised yet highly conserved surfaces, many of which contain clinically significant mutations. The xProtCAS tool is available as open-source Python software and an interactive web server.

Blind assessment of monomeric AlphaFold2 protein structure models with experimental NMR data

Recent advances in molecular modeling of protein structures are changing the field of structural biology. AlphaFold-2 (AF2), an AI system developed by DeepMind, Inc., utilizes attention-based deep learning to predict models of protein structures with high accuracy relative to structures determined by X-ray crystallography and cryo-electron microscopy (cryoEM). Comparing AF2 models to structures determined using solution NMR data, both high similarities and distinct differences have been observed. Since AF2 was trained on X-ray crystal and cryoEM structures, we assessed how accurately AF2 can model small, monomeric, solution protein NMR structures which (i) were not used in the AF2 training data set, and (ii) did not have homologous structures in the Protein Data Bank at the time of AF2 training. We identified nine open-source protein NMR data sets for such “blind” targets, including chemical shift, raw NMR FID data, NOESY peak lists, and (for 1 case) 15N-1H residual dipolar coupling data. For these nine small (70–108 residues) monomeric proteins, we generated AF2 prediction models and assessed how well these models fit to these experimental NMR data, using several well-established NMR structure validation tools. In most of these cases, the AF2 models fit the NMR data nearly as well, or sometimes better than, the corresponding NMR structure models previously deposited in the Protein Data Bank. These results provide benchmark NMR data for assessing new NMR data analysis and protein structure prediction methods. They also document the potential for using AF2 as a guiding tool in protein NMR data analysis, and more generally for hypothesis generation in structural biology research.

AlphaFold2 Update and Perspectives

Access to the three-dimensional (3D) structural information of macromolecules is of major interest in both fundamental and applied research. Obtaining this experimental data can be complex, time consuming, and costly. Therefore, in silico computational approaches are an alternative of interest, and sometimes present a unique option. In this context, the Protein Structure Prediction method AlphaFold2 represented a revolutionary advance in structural bioinformatics. Named method of the year in 2021, and widely distributed by DeepMind and EBI, it was thought at this time that protein-folding issues had been resolved. However, the reality is slightly more complex. Due to a lack of input experimental data, related to crystallographic challenges, some targets have remained highly challenging or not feasible. This perspective exercise, dedicated to a non-expert audience, discusses and correctly places AlphaFold2 methodology in its context and, above all, highlights its use, limitations, and opportunities. After a review of the interest in the 3D structure and of the previous methods used in the field, AF2 is brought into its historical context. Its spatial interests are detailed before presenting precise quantifications showing some limitations of this approach and finishing with the perspectives in the field.

Atomic protein structure refinement using all-atom graph representations and SE(3)-equivariant graph transformer.

The state-of-art protein structure prediction methods such as AlphaFold are being widely used to predict structures of uncharacterized proteins in biomedical research. There is a significant need to further improve the quality and nativeness of the predicted structures to enhance their usability. In this work, we develop ATOMRefine, a deep learning-based, end-to-end, all-atom protein structural model refinement method. It uses a SE(3)-equivariant graph transformer network to directly refine protein atomic coordinates in a predicted tertiary structure represented as a molecular graph. The method is first trained and tested on the structural models in AlphaFoldDB whose experimental structures are known, and then blindly tested on 69 CASP14 regular targets and 7 CASP14 refinement targets. ATOMRefine improves the quality of both backbone atoms and all-atom conformation of the initial structural models generated by AlphaFold. It also performs better than two state-of-the-art refinement methods in multiple evaluation metrics including an all-atom model quality score-the MolProbity score based on the analysis of all-atom contacts, bond length, atom clashes, torsion angles, and side-chain rotamers. As ATOMRefine can refine a protein structure quickly, it provides a viable, fast solution for improving protein geometry and fixing structural errors of predicted structures through direct coordinate refinement. The source code of ATOMRefine is available in the GitHub repository (https://github.com/BioinfoMachineLearning/ATOMRefine). All the required data for training and testing are available at https://doi.org/10.5281/zenodo.6944368.

Pareto Dominance Archive and Coordinated Selection Strategy-Based Many-Objective Optimizer for Protein Structure Prediction.

Protein structure prediction (PSP) is predicting the three-dimensional of protein from its amino acid sequence only based on the information hidden in the protein sequence. One of the efficient tools to describe this information is protein energy functions. Despite the advancements in biology and computer science, PSP is still a challenging problem due to its large protein conformation space and inaccurate energy functions. In this study, PSP is treated as a many-objective optimization problem and four conflicting energy functions are used as different objectives to be optimized. A novel Pareto-dominance-archive and Coordinated-selection-strategy-based Many-objective-optimizer (PCM) is proposed to perform the conformation search. In it, convergence and diversity-based selection metrics are used to enable PCM to find near-native proteins with well-distributed energy values, while a Pareto-dominance-based archive is proposed to save more potential conformations that can guide the search to more promising conformation areas. The experimental results on thirty-four benchmark proteins demonstrate the significant superiority of PCM in comparison with other single, multiple, and many-objective evolutionary algorithms. Additionally, the inherent characteristics of iterative search of PCM can also give more insights into the dynamic progress of protein folding besides the final predicted static tertiary structure. All these confirm that PCM is a fast, easy-to-use, and fruitful solution generation method for PSP.

AlphaFold predicts novel human proteins with knots.

The fact that proteins can have their chain formed in a knot is known for almost 30 years. However, as they are not common, only a fraction of such proteins is available in the Protein Data Bank. It was not possible to assess their importance and versatility up until now because we did not have access to the whole proteome of an organism, let alone a human one. The arrival of efficient machine learning methods for protein structure prediction, such as AlphaFold and RoseTTaFold, changed that. We analyzed all proteins from the human proteome (over 20,000) determined with AlphaFold in search for knots and found them in less than 2% of the structures. Using a variety of methods, including homolog search, clustering, quality assessment and visual inspection, we determined the nature of each of the knotted structures and classified it as either knotted, potentially knotted, or an artifact, and deposited all of them in a database available at: https://knotprot.cent.uw.edu.pl/alphafold. Overall, we found 51 credible knotted proteins (0.2% of human proteome). The set of potentially knotted structures includes a new complex type of a knot not reported in proteins yet. That knot type, denoted 63 in mathematical notation, would necessitate a more complex folding path than any knotted protein characterized to date. This article is protected by copyright. All rights reserved.

Protein Secondary Structure Prediction Using Cascaded Feature Learning Model

The protein secondary structure prediction (PSSP) is pivotal for predicting tertiary structure, which is proliferating in demand for drug design and development. Further, it can be used to learn different protein functions. Although there are many computational methods for protein structure prediction, none of them have succeeded 100% in solving the protein structure prediction problem. This paper proposes a deep learning model, namely Cascaded Feature Learning Model (CFLM) that uses a multi-stage transfer learning approach based on the Residual Dense Network (RDN) for predicting protein secondary structure. The model is trained with different protein datasets at various levels of transfer learning and accepts selected protein features such as solvent accessibility, physicochemical properties, PSSM (position-specific scoring matrix), and PSFM (position-specific frequency matrix) as input. The Q3 and Q8 accuracy obtained on CASP 13 and 14 benchmark datasets with the proposed approach are 91.23% and 91.45%, and 76.83% and 78.04%, respectively. The performance of CFLM is also compared with some of the recent PSSP approaches and it is observed that the proposed approach improves the prediction accuracy by 2.8% in terms of Q3 and 1.81% in terms of Q8 in the CASP 14 dataset.

High-Risk Pedigree Study Identifies LRBA (rs62346982) as a Likely Predisposition Variant for Prostate Cancer.

There is evidence for contribution of inherited factors to prostate cancer, and more specifically to lethal prostate cancer, but few responsible genes/variants have been identified. We examined genetic sequence data for 51 affected cousin pairs who each died from prostate cancer and who were members of high-risk prostate cancer pedigrees in order to identify rare variants shared by the cousins as candidate predisposition variants. Candidate variants were tested for association with prostate cancer risk in UK Biobank data. Candidate variants were also assayed in 1195 additional sampled Utah prostate cancer cases. We used 3D protein structure prediction methods to analyze structural changes and provide insights into mechanisms of pathogenicity. Almost 4000 rare (<0.005) variants were identified as shared in the 51 affected cousin pairs. One candidate variant was also significantly associated with prostate cancer risk among the 840 variants with data in UK Biobank, in the gene LRBA (p = 3.2 × 10-5; OR = 2.09). The rare risk variant in LRBA was observed to segregate in five pedigrees. The overall predicted structures of the mutant protein do not show any significant overall changes upon mutation, but the mutated structure loses a helical structure for the two residues after the mutation. This unique analysis of closely related individuals with lethal prostate cancer, who were members of high-risk prostate cancer pedigrees, has identified a strong set of candidate predisposition variants which should be pursued in independent studies. Validation data for a subset of the candidates identified are presented, with strong evidence for a rare variant in LRBA.

AFTM: a database of transmembrane regions in the human proteome predicted by AlphaFold.

Transmembrane proteins (TMPs), with diverse cellular functions, are difficult targets for structural determination. Predictions of TMPs and the locations of transmembrane segments using computational methods could be unreliable due to the potential for false positives and false negatives and show inconsistencies across different programs. Recent advances in protein structure prediction methods have made it possible to identify TMPs and their membrane-spanning regions using high-quality structural models. We developed the AlphaFold Transmembrane proteins (AFTM) database of candidate human TMPs by identifying transmembrane regions in AlphaFold structural models of human proteins and their domains using the positioning of proteins in membranes, version 3 program, followed by automatic corrections inspired by manual analysis of the results. We compared our results to annotations from the UniProt database and the Human Transmembrane Proteome (HTP) database. While AFTM did not identify transmembrane regions in some single-pass TMPs, it identified more transmembrane regions for multipass TMPs than UniProt and HTP. AFTM also showed more consistent results with experimental structures, as benchmarked against the Protein Data Bank Transmembrane proteins (PDBTM) database. In addition, some proteins previously annotated as TMPs were suggested to be non-TMPs by AFTM. We report the results of AFTM together with those of UniProt, HTP, TmAlphaFold, PDBTM and Membranome in the online AFTM database compiled as a comprehensive resource of candidate human TMPs with structural models. Database URL http://conglab.swmed.edu/AFTM.

Recent Advances in Biomolecular Structure Modeling and Validation using Deep Learning

Since its appearance in 2021, Alphafold2, an accurate protein structure prediction method, has quickly been adopted and substantially impacted biology research, particularly in the field of structural biology. In this review, we discuss recent advancements in the development of computational structural modeling tools related to Alphafold2. Additionally, we discuss deep learning-based structure modeling methods for cryo-electron microscopy.

Real-to-bin conversion for protein residue distances

Protein Structure Prediction (PSP) has achieved significant progress lately. Prediction of inter-residue distances by machine learning and their exploitation during the conformational search is largely among the critical factors behind the progress. Real values than bin probabilities could more naturally represent inter-residue distances, while the latter, via spline curves more naturally helps obtain differentiable objective functions than the former. Consequently, PSP methods that exploit predicted binned distances perform better than those that exploit predicted real-valued distances. To leverage the advantage of bin probabilities in getting differentiable objective functions, in this work, we propose techniques to convert real-valued distances into distance bin probabilities. Using standard benchmark proteins, we then show that our real-to-bin converted distances help PSP methods obtain three-dimensional structures with 4%–16% better root mean squared deviation (RMSD), template modeling score (TM-Score), and global distance test (GDT) values than existing similar PSP methods. Our proposed PSP method is named real to bin (R2B) inter-residue distance predictor, and its code is available from https://gitlab.com/mahnewton/r2b.

High-accuracy protein model quality assessment using attention graph neural networks.

Great improvement has been brought to protein tertiary structure prediction through deep learning. It is important but very challenging to accurately rank and score decoy structures predicted by different models. CASP14 results show that existing quality assessment (QA) approaches lag behind the development of protein structure prediction methods, where almost all existing QA models degrade in accuracy when the target is a decoy of high quality. How to give an accurate assessment to high-accuracy decoys is particularly useful with the available of accurate structure prediction methods. Here we propose a fast and effective single-model QA method, QATEN, which can evaluate decoys only by their topological characteristics and atomic types. Our model uses graph neural networks and attention mechanisms to evaluate global and amino acid level scores, and uses specific loss functions to constrain the network to focus more on high-precision decoys and protein domains. On the CASP14 evaluation decoys, QATEN performs better than other QA models under all correlation coefficients when targeting average LDDT. QATEN shows promising performance when considering only high-accuracy decoys. Compared to the embedded evaluation modules of predicted ${C}_{\alpha^{-}} RMSD$ (pRMSD) in RosettaFold and predicted LDDT (pLDDT) in AlphaFold2, QATEN is complementary and capable of achieving better evaluation on some decoy structures generated by AlphaFold2 and RosettaFold. These results suggest that the new QATEN approach can be used as a reliable independent assessment algorithm for high-accuracy protein structure decoys.

PyScoMotif: discovery of similar 3D structural motifs across proteins.

The fast and accurate detection of similar geometrical arrangements of protein residues, known as 3D structural motifs, is highly relevant for many applications such as binding region and catalytic site detection, drug discovery and structure conservation analyses. With the recent publication of new protein structure prediction methods, the number of available protein structures is exploding, which makes efficient and easy-to-use tools for identifying 3D structural motifs essential. We present an open-source Python package that enables the search for both exact and mutated motifs with position-specific residue substitutions. The tool is efficient, flexible, accurate, and suitable to run both on computer clusters and personal laptops. Two successful applications of pyScoMotif for catalytic site identification are showcased. The pyScoMotif package can be installed from the PyPI repository and is also available at https://github.com/3BioCompBio/pyScoMotif. It is free to use for non-commercial purposes.

Mapping the FtsQBL divisome components in bacterial NTD pathogens as potential drug targets.

Cytokinesis is an essential process in bacterial cell division, and it involves more than 25 essential/non-essential cell division proteins that form a protein complex known as a divisome. Central to the divisome are the proteins FtsB and FtsL binding to FtsQ to form a complex FtsQBL, which helps link the early proteins with late proteins. The FtsQBL complex is highly conserved as a component across bacteria. Pathogens like Vibrio cholerae, Mycobacterium ulcerans, Mycobacterium leprae, and Chlamydia trachomatis are the causative agents of the bacterial Neglected Tropical Diseases Cholera, Buruli ulcer, Leprosy, and Trachoma, respectively, some of which seemingly lack known homologs for some of the FtsQBL complex proteins. In the absence of experimental characterization, either due to insufficient resources or the massive increase in novel sequences generated from genomics, functional annotation is traditionally inferred by sequence similarity to a known homolog. With the advent of accurate protein structure prediction methods, features both at the fold level and at the protein interaction level can be used to identify orthologs that cannot be unambiguously identified using sequence similarity methods. Using the FtsQBL complex proteins as a case study, we report potential remote homologs using Profile Hidden Markov models and structures predicted using AlphaFold. Predicted ortholog structures show conformational similarity with corresponding E. coli proteins irrespective of their level of sequence similarity. Alphafold multimer was used to characterize remote homologs as FtsB or FtsL, when they were not sufficiently distinguishable at both the sequence or structure level, as their interactions with FtsQ and FtsW play a crucial role in their function. The structures were then analyzed to identify functionally critical regions of the proteins consistent with their homologs and delineate regions potentially useful for inhibitor discovery.

Towards de novo design of transmembrane α-helical assemblies using structural modelling and molecular dynamics simulation.

Computational de novo protein design involves iterative processes consisting of amino acid sequence design, structural modelling and scoring, and design validation by synthesis and experimental characterisation. Recent advances in protein structure prediction and modelling methods have enabled the highly efficient and accurate design of water-soluble proteins. However, the design of membrane proteins remains a major challenge. To advance membrane protein design, considering the higher complexity of membrane protein folding, stability, and dynamic interactions between water, ions, lipids, and proteins is an important task. For introducing explicit solvents and membranes to these design methods, all-atom molecular dynamics (MD) simulations of designed proteins provide useful information that cannot be obtained experimentally. In this review, we first describe two major approaches to designing transmembrane α-helical assemblies, consensus and de novo design. We further illustrate recent MD studies of membrane protein folding related to protein design, as well as advanced treatments in molecular models and conformational sampling techniques in the simulations. Finally, we discuss the possibility to introduce MD simulations after the existing static modelling and screening of design decoys as an additional step for refinement of the design, which considers membrane protein folding dynamics and interactions with explicit membranes.