Inter-residue Distances Research Articles

New Labeling Methods for Deep Learning Real-Valued Inter-Residue Distance Prediction.

Much of the recent success in protein structure prediction has been a result of accurate protein contact prediction-a binary classification problem. Dozens of methods, built from various types of machine learning and deep learning algorithms, have been published over the last two decades for predicting contacts. Recently, many groups, including Google DeepMind, have demonstrated that reformulating the problem as a multi-class classification problem is a more promising direction to pursue. As an alternative approach, we recently proposed real-valued distance predictions, formulating the problem as a regression problem. The nuances of protein 3D structures make this formulation appropriate, allowing predictions to reflect inter-residue distances in nature. Despite these promises, the accurate prediction of real-valued distances remains relatively unexplored; possibly due to classification being better suited to machine and deep learning algorithms. Can regression methods be designed to predict real-valued distances as precise as binary contacts? To investigate this, we propose multiple novel methods of input label engineering, which is different from feature engineering, with the goal of optimizing the distribution of distances to cater to the loss function of the deep-learning model. Since an important utility of predicted contacts or distances is to build three-dimensional models, we also tested if predicted distances can reconstruct more accurate models than contacts. Our results demonstrate, for the first time, that deep learning methods for real-valued protein distance prediction can deliver distances as precise as binary classification methods. When using an optimal distance transformation function on the standard PSICOV dataset consisting of 150 representative proteins, the precision of 'top-all' long-range contacts improves from 60.9% to 61.4% when predicting real-valued distances instead of contacts. When building three-dimensional models we observed an average TM-score increase from 0.61 to 0.72, highlighting the advantage of predicting real-valued distances.

Biased agonists differentially modulate the receptor conformation ensembles in Angiotensin II type 1 receptor

The structural features that contribute to the efficacy of biased agonists targeting G protein-coupled receptors (GPCRs) towards G proteins or β-arrestin (β-arr) signaling pathways is nebulous, although such knowledge is critical in designing biased ligands. The dynamics of the agonist-GPCR complex is one of the critical factors in determining agonist bias. Angiotensin II type I receptor (AT1R) is an ideal model system to study the molecular basis of bias since it has multiple β-arr2 and Gq protein biased agonists as well as experimentally solved three dimensional structures. Using Molecular Dynamics (MD) simulations for the Angiotensin II type I receptor (AT1R) bound to ten different agonists, we infer that the agonist bound receptor samples conformations with different relative weights, from both the inactive and active state ensembles of the receptor. This concept is perhaps extensible to other class A GPCRs. Such a weighted mixed ensemble recapitulates the inter-residue distance distributions measured for different agonists bound AT1R using DEER experiments. The ratio of the calculated relative strength of the allosteric communication to β-arr2 vs Gq coupling sites scale similarly to the experimentally measured bias factors. Analysis of the inter-residue distance distributions of the activation microswitches involved in class A GPCR activation suggests that β-arr2 biased agonists turn on different combination of microswitches with different relative strengths of activation. We put forth a model that activation microswitches behave like rheostats that tune the relative efficacy of the biased agonists toward the two signaling pathways. Finally, based on our data we propose that the agonist specific residue contacts in the binding site elicit a combinatorial response in the microswitches that in turn differentially modulate the receptor conformation ensembles resulting in differences in coupling to Gq and β-arrestin.

Structural analogue-based protein structure domain assembly assisted by deep learning.

With the breakthrough of AlphaFold2, the protein structure prediction problem has made remarkable progress through deep learning end-to-end techniques, in which correct folds could be built for nearly all single-domain proteins. However, the full-chain modelling appears to be lower on average accuracy than that for the constituent domains and requires higher demand on computing hardware, indicating the performance of full-chain modelling still needs to be improved. In this study, we investigate whether the predicted accuracy of the full-chain model can be further improved by domain assembly assisted by deep learning. In this article, we developed a structural analogue-based protein structure domain assembly method assisted by deep learning, named SADA. In SADA, a multi-domain protein structure database was constructed for the full-chain analogue detection using individual domain models. Starting from the initial model constructed from the analogue, the domain assembly simulation was performed to generate the full-chain model through a two-stage differential evolution algorithm guided by the energy function with an inter-residue distance potential predicted by deep learning. SADA was compared with the state-of-the-art domain assembly methods on 356 benchmark proteins, and the average TM-score of SADA models is 8.1% and 27.0% higher than that of DEMO and AIDA, respectively. We also assembled 293 human multi-domain proteins, where the average TM-score of the full-chain model after the assembly by SADA is 1.1% higher than that of the model by AlphaFold2. To conclude, we find that the domains often interact in the similar way in the quaternary orientations if the domains have similar tertiary structures. Furthermore, homologous templates and structural analogues are complementary for multi-domain protein full-chain modelling. http://zhanglab-bioinf.com/SADA. Supplementary data are available at Bioinformatics online.

Evolution of Stronger SARS-CoV-2 Variants as Revealed Through the Lens of Molecular Dynamics Simulations.

Using molecular dynamics simulations, the protein–protein interactions of the receptor-binding domain of the wild-type and seven variants of the severe acute respiratory syndrome coronavirus 2 spike protein and the peptidase domain of human angiotensin-converting enzyme 2 were investigated. These variants are alpha, beta, gamma, delta, eta, kappa, and omicron. Using 100 ns simulation data, the residue interaction networks at the protein–protein interface were identified. Also, the impact of mutations on essential protein dynamics, backbone flexibility, and interaction energy of the simulated protein–protein complexes were studied. The protein–protein interface for the wild-type, delta, and omicron variants contained several stronger interactions, while the alpha, beta, gamma, eta, and kappa variants exhibited an opposite scenario as evident from the analysis of the inter-residue interaction distances and pair-wise interaction energies. The study reveals that two distinct residue networks at the central and right contact regions forge stronger binding affinity between the protein partners. The study provides a molecular-level insight into how enhanced transmissibility and infectivity by delta and omicron variants are most likely tied to a handful of interacting residues at the binding interface, which could potentially be utilized for future antibody constructs and structure-based antiviral drug design. Supplementary InformationThe online version contains supplementary material available at 10.1007/s10930-022-10065-6.

Multi-head attention-based U-Nets for predicting protein domain boundaries using 1D sequence features and 2D distance maps

The information about the domain architecture of proteins is useful for studying protein structure and function. However, accurate prediction of protein domain boundaries (i.e., sequence regions separating two domains) from sequence remains a significant challenge. In this work, we develop a deep learning method based on multi-head U-Nets (called DistDom) to predict protein domain boundaries utilizing 1D sequence features and predicted 2D inter-residue distance map as input. The 1D features contain the evolutionary and physicochemical information of protein sequences, whereas the 2D distance map includes the structural information of proteins that was rarely used in domain boundary prediction before. The 1D and 2D features are processed by the 1D and 2D U-Nets respectively to generate hidden features. The hidden features are then used by the multi-head attention to predict the probability of each residue of a protein being in a domain boundary, leveraging both local and global information in the features. The residue-level domain boundary predictions can be used to classify proteins as single-domain or multi-domain proteins. It classifies the CASP14 single-domain and multi-domain targets at the accuracy of 75.9%, 13.28% more accurate than the state-of-the-art method. Tested on the CASP14 multi-domain protein targets with expert annotated domain boundaries, the average per-target F1 measure score of the domain boundary prediction by DistDom is 0.263, 29.56% higher than the state-of-the-art method.

A stacked meta-ensemble for protein inter-residue distance prediction

Predicted inter-residue distances are a key behind recent success in high quality protein structure prediction (PSP). However, prediction of both short and long distance values together is challenging. Consequently, predicted short distances are mostly used by existing PSP methods. In this paper, we use a stacked meta-ensemble method to combine deep learning models trained for different ranges of real-valued distances. On five benchmark sets of proteins, our proposed inter-residue distance prediction method improves mean Local Distance Different Test (LDDT) scores at least by 5% over existing such methods. Moreover, using a real-valued distance based conformational search algorithm, we also show that predicted long distances help obtain significantly better protein conformations than when only predicted short distances are used. Our method is named meta-ensemble for distance prediction (MDP) and its program is available from https://gitlab.com/mahnewton/mdp.

Correlation-Based Feature Selection to Identify Functional Dynamics in Proteins.

To interpret molecular dynamics simulations of biomolecular systems, systematic dimensionality reduction methods are commonly employed. Among others, this includes principal component analysis (PCA) and time-lagged independent component analysis (TICA), which aim to maximize the variance and the time scale of the first components, respectively. A crucial first step of such an analysis is the identification of suitable and relevant input coordinates (the so-called features), such as backbone dihedral angles and interresidue distances. As typically only a small subset of those coordinates is involved in a specific biomolecular process, it is important to discard the remaining uncorrelated motions or weakly correlated noise coordinates. This is because they may exhibit large amplitudes or long time scales and therefore will be erroneously considered important by PCA and TICA, respectively. To discriminate collective motions underlying functional dynamics from uncorrelated motions, the correlation matrix of the input coordinates is block-diagonalized by a clustering method. This strategy avoids possible bias due to presumed functional observables and conformational states or variation principles that maximize variance or time scales. Considering several linear and nonlinear correlation measures and various clustering algorithms, it is shown that the combination of linear correlation and the Leiden community detection algorithm yields excellent results for all considered model systems. These include the functional motion of T4 lysozyme to demonstrate the successful identification of collective motion, as well as the folding of the villin headpiece to highlight the physical interpretation of the correlated motions in terms of a functional mechanism.

Simultaneous prediction of antibody backbone and side-chain conformations with deep learning.

Antibody engineering is becoming increasingly popular in medicine for the development of diagnostics and immunotherapies. Antibody function relies largely on the recognition and binding of antigenic epitopes via the loops in the complementarity determining regions. Hence, accurate high-resolution modeling of these loops is essential for effective antibody engineering and design. Deep learning methods have previously been shown to effectively predict antibody backbone structures described as a set of inter-residue distances and orientations. However, antigen binding is also dependent on the specific conformations of surface side-chains. To address this shortcoming, we created DeepSCAb: a deep learning method that predicts inter-residue geometries as well as side-chain dihedrals of the antibody variable fragment. The network requires only sequence as input, rendering it particularly useful for antibodies without any known backbone conformations. Rotamer predictions use an interpretable self-attention layer, which learns to identify structurally conserved anchor positions across several species. We evaluate the performance of the model for discriminating near-native structures from sets of decoys and find that DeepSCAb outperforms similar methods lacking side-chain context. When compared to alternative rotamer repacking methods, which require an input backbone structure, DeepSCAb predicts side-chain conformations competitively. Our findings suggest that DeepSCAb improves antibody structure prediction with accurate side-chain modeling and is adaptable to applications in docking of antibody-antigen complexes and design of new therapeutic antibody sequences.

Cooperative Protein Allosteric Transition Mediated by a Fluctuating Transmission Network

Allosteric communication between distant protein sites represents a key mechanism of biomolecular regulation and signal transduction. Compared to other processes such as protein folding, however, the dynamical evolution of allosteric transitions is still not well understood. As an example of allosteric coupling between distant protein regions, we consider the global open-closed motion of the two domains of T4 lysozyme, which is triggered by local motion in the hinge region. Combining extensive molecular dynamics simulations with a correlation analysis of interresidue contacts, we identify a network of interresidue distances that move in a concerted manner. The cooperative process originates from a cogwheel-like motion of the hydrophobic core in the hinge region, which constitutes an evolutionary conserved and flexible transmission network. Through rigid contacts and the protein backbone, the small local changes of the hydrophobic core are passed on to the distant terminal domains and lead to the emergence of a rare global conformational transition. As in an Ising-type model, the cooperativity of the allosteric transition can be explained via the interaction of local fluctuations.

Unique SMYD5 Structure Revealed by AlphaFold Correlates with Its Functional Divergence

SMYD5 belongs to a special class of protein lysine methyltransferases with an MYND (Myeloid-Nervy-DEAF1) domain inserted into a SET (Suppressor of variegation, Enhancer of Zeste, Trithorax) domain. Despite recent advances in its functional characterization, the lack of the crystal structure has hindered our understanding of the structure-and-function relationships of this most unique member of the SMYD protein family. Here, we demonstrate the reliability of using AlphaFold structures for understanding the structure and function of SMYD5 by comparing the AlphaFold structures to the known crystal structures of SMYD proteins, using an inter-residue distance maps-based metric. We found that the AlphaFold confidence scores are inversely associated with the refined B-factors and can serve as a structural indicator of conformational flexibility. We also found that the N-terminal sequence of SMYD5, predicted to be a mitochondrial targeting signal, contains a novel non-classical nuclear localization signal. This sequence is structurally flexible and does not have a well-defined conformation, which might facilitate its recognition for SMYD5’s cytonuclear transport. The structure of SMYD5 is unique in many aspects. The “crab”-like structure with a large negatively charged cleft provides a potential binding site for basic molecules such as protamines. The less positively charged MYND domain is associated with the undetectable DNA-binding ability. The most surprising feature is an incomplete target lysine access channel that lacks the evolutionarily conserved tri-aromatic arrangement, being associated with the low H3/H4 catalytic activity. This study expands our understanding of the SMYD protein family from a classical two-lobed structure to a structure of its own kind, being as a fundamental determinant of its functional divergence.

Protein language-model embeddings for fast, accurate, and alignment-free protein structure prediction

Advanced protein structure prediction requires evolutionary information from multiple sequence alignments (MSAs) from evolutionary couplings that are not always available. Artificial intelligence (AI)-based predictions inputting only single sequences are faster but so inaccurate as to render speed irrelevant. Here, we described a competitive prediction of inter-residue distances (2D structure) exclusively inputting embeddings from pre-trained protein language models (pLMs), namely ProtT5, from single sequences into a convolutional neural network (CNN) with relatively few layers. The major advance used the ProtT5 attention heads. Our new method, EMBER2, which never requires any MSAs, performed similarly to other methods that fully rely on co-evolution. Although clearly not reaching AlphaFold2, our leaner solution came somehow close at substantially lower costs. By generating protein-specific rather than family-averaged predictions, EMBER2 might better capture some features of particular protein structures. Results from using protein engineering and deep mutational scanning (DMS) experiments provided at least a proof of principle for such a speculation.

Enhancing protein contact map prediction accuracy via ensembles of inter-residue distance predictors

Protein contact maps capture coevolutionary interactions between amino acid residue pairs that are spatially within certain proximity threshold. Predicted contact maps are used in many protein related problems that include drug design, protein design, protein function prediction, and protein structure prediction. Contact map prediction has achieved significant progress lately but still further challenges remain with prediction of contacts between residues that are separated in the amino acid residue sequence by large numbers of other residues. In this paper, with experimental results on 5 standard benchmark datasets that include membrane proteins, we show that contact map prediction could be significantly enhanced by using ensembles of various state-of-the-art short distance predictors and then by converting predicted distances into contact probabilities. Our program along with its data is available from https://gitlab.com/mahnewton/ecp.

Inter-Residue Distance Prediction From Duet Deep Learning Models.

Residue distance prediction from the sequence is critical for many biological applications such as protein structure reconstruction, protein–protein interaction prediction, and protein design. However, prediction of fine-grained distances between residues with long sequence separations still remains challenging. In this study, we propose DuetDis, a method based on duet feature sets and deep residual network with squeeze-and-excitation (SE), for protein inter-residue distance prediction. DuetDis embraces the ability to learn and fuse features directly or indirectly extracted from the whole-genome/metagenomic databases and, therefore, minimize the information loss through ensembling models trained on different feature sets. We evaluate DuetDis and 11 widely used peer methods on a large-scale test set (610 proteins chains). The experimental results suggest that 1) prediction results from different feature sets show obvious differences; 2) ensembling different feature sets can improve the prediction performance; 3) high-quality multiple sequence alignment (MSA) used for both training and testing can greatly improve the prediction performance; and 4) DuetDis is more accurate than peer methods for the overall prediction, more reliable in terms of model prediction score, and more robust against shallow multiple sequence alignment (MSA).

Decoding an Amino Acid Sequence to Extract Information on Protein Folding.

Protein folding is a complicated phenomenon including various time scales (μs to several s), and various structural indices are required to analyze it. The methodologies used to study this phenomenon also have a wide variety and employ various experimental and computational techniques. Thus, a simple speculation does not serve to understand the folding mechanism of a protein. In the present review, we discuss the recent studies conducted by the author and their colleagues to decode amino acid sequences to obtain information on protein folding. We investigate globin-like proteins, ferredoxin-like fold proteins, IgG-like beta-sandwich fold proteins, lysozyme-like fold proteins and β-trefoil-like fold proteins. Our techniques are based on statistics relating to the inter-residue average distance, and our studies performed so far indicate that the information obtained from these analyses includes data on the protein folding mechanism. The relationships between our results and the actual protein folding phenomena are also discussed.

Multifunctions of α‐Synuclein Explained by Its Dynamic Heterogeneous Conformations with a Hierarchy of Transition Times

α‐Synuclein plays essential roles in synaptic vesicle homeostasis and neurotransmitter release through the interaction with membrane, while α‐synuclein aggregates have been associated with several neurodegenerative diseases, especially Parkinson’s disease. The different roles of α‐synuclein could be attributed to the heterogenous conformations adopted by this intrinsically disordered protein. To demystify the physiological functions of α‐synuclein and understand its pathogenic mechanism of Parkinson’s disease, characterizing the monomeric structural ensemble and identifying the aggregation‐prone and non‐aggregation‐prone structures are of particular importance. Here, we use inter‐residue distance distributions derived from time‐resolved FRET experiments as constraints to guide discrete molecular dynamics simulations of α‐synuclein monomer. We explore the conformational space of α‐synuclein and verify the generated conformational ensemble by additional experiments including far‐UV circular dichroism spectrum and cross‐linking mass spectrometry. We find that some conformational states of α‐synuclein are surprisingly stable displaying dynamic transitions less than milliseconds. A comprehensive analysis of the conformational ensemble uncovers important structural features and potential conformations that are critical to stabilize the monomeric state or induce different oligomerization pathways.

DISTEMA: distance map-based estimation of single protein model accuracy with attentive 2D convolutional neural network

BackgroundEstimation of the accuracy (quality) of protein structural models is important for both prediction and use of protein structural models. Deep learning methods have been used to integrate protein structure features to predict the quality of protein models. Inter-residue distances are key information for predicting protein’s tertiary structures and therefore have good potentials to predict the quality of protein structural models. However, few methods have been developed to fully take advantage of predicted inter-residue distance maps to estimate the accuracy of a single protein structural model.ResultWe developed an attentive 2D convolutional neural network (CNN) with channel-wise attention to take only a raw difference map between the inter-residue distance map calculated from a single protein model and the distance map predicted from the protein sequence as input to predict the quality of the model. The network comprises multiple convolutional layers, batch normalization layers, dense layers, and Squeeze-and-Excitation blocks with attention to automatically extract features relevant to protein model quality from the raw input without using any expert-curated features. We evaluated DISTEMA’s capability of selecting the best models for CASP13 targets in terms of ranking loss of GDT-TS score. The ranking loss of DISTEMA is 0.079, lower than several state-of-the-art single-model quality assessment methods.ConclusionThis work demonstrates that using raw inter-residue distance information with deep learning can predict the quality of protein structural models reasonably well. DISTEMA is freely at https://github.com/jianlin-cheng/DISTEMA

Accelerating Protein Folding Molecular Dynamics Using Inter-Residue Distances from Machine Learning Servers.

Recently, predicting the native structures of proteins has become possible using computational molecular physics (CMP)─physics-based force fields sampled with proper statistics─but only for small proteins. Algorithms with better scaling are needed. We describe ML x MELD x MD, a molecular dynamics (MD) method that inputs residue contacts derived from machine learning (ML) servers into MELD, a Bayesian accelerator that preserves detailed-balance statistics. Contacts are derived from trRosetta-predicted distance histograms (distograms) and are integrated into MELD's atomistic MD as spatial restraints through parametrized potential functions. In the CASP14 blind prediction event, ML x MELD x MD predicted 13 native structures to better than 4.5 Å error, including for 10 proteins in the range of 115-250 amino acids long. Also, the scaling of simulation time vs protein length is much better than unguided MD: tsim ∼ e0.023N for ML x MELD x MD vs tsim ∼ e0.168N for MD alone. This shows how machine learning information can be leveraged to advance physics-based modeling of proteins.

Enhancing protein inter-residue real distance prediction by scrutinising deep learning models

Protein structure prediction (PSP) has achieved significant progress lately via prediction of inter-residue distances using deep learning models and exploitation of the predictions during conformational search. In this context, prediction of large inter-residue distances and also prediction of distances between residues separated largely in the protein sequence remain challenging. To deal with these challenges, state-of-the-art inter-residue distance prediction algorithms have used large sets of coevolutionary and non-coevolutionary features. In this paper, we argue that the more the types of features used, the more the kinds of noises introduced and then the deep learning model has to overcome the noises to improve the accuracy of the predictions. Also, multiple features capturing similar underlying characteristics might not necessarily have significantly better cumulative effect. So we scrutinise the feature space to reduce the types of features to be used, but at the same time, we strive to improve the prediction accuracy. Consequently, for inter-residue real distance prediction, in this paper, we propose a deep learning model named scrutinised distance predictor (SDP), which uses only 2 coevolutionary and 3 non-coevolutionary features. On several sets of benchmark proteins, our proposed SDP method improves mean Local Distance Different Test (LDDT) scores at least by 10% over existing state-of-the-art methods. The SDP program along with its data is available from the website https://gitlab.com/mahnewton/sdp.

De novo protein design by deep network hallucination.

There has been considerable recent progress in protein structure prediction using deep neural networks to predict inter-residue distances from amino acid sequences1-3. Here we investigate whether the information captured by such networks is sufficiently rich to generate new folded proteins with sequences unrelated to those of the naturally occurring proteins used in training the models. We generate random amino acid sequences, and input them into the trRosetta structure prediction network to predict starting residue-residue distance maps, which, as expected, are quite featureless. We then carry out Monte Carlo sampling in amino acid sequence space, optimizing the contrast (Kullback-Leibler divergence) between the inter-residuedistance distributions predicted by the network and background distributions averaged over all proteins. Optimization from different random starting points resulted in novel proteins spanning a wide range of sequences and predicted structures. We obtained synthetic genes encoding 129 of the network-'hallucinated' sequences, and expressed and purified the proteins in Escherichia coli; 27 of the proteins yielded monodisperse species with circular dichroism spectra consistent with the hallucinated structures. We determined the three-dimensionalstructures of three of the hallucinated proteins, two by X-ray crystallography and one by NMR, and these closely matched the hallucinated models. Thus, deep networks trained to predict native protein structures from their sequences can be inverted to design new proteins, and such networks and methods should contribute alongside traditional physics-based models to the de novo design of proteins with new functions.

Toward the assessment of predicted inter-residue distance.

Significant progress has been achieved in distance-based protein folding, due to improved prediction of inter-residue distance by deep learning. Many efforts are thus made to improve distance prediction in recent years. However, it remains unknown what is the best way of objectively assessing the accuracy of predicted distance. A total of 19 metrics were proposed to measure the accuracy of predicted distance. These metrics were discussed and compared quantitatively on three benchmark datasets, with distance and structure models predicted by the trRosetta pipeline. The experiments show that a few metrics, such as distance precision, have a high correlation with the model accuracy measure TM-score (Pearson's correlation coefficient >0.7). In addition, the metrics are applied to rank the distance prediction groups in CASP14. The ranking by our metrics coincides largely with the official version. These data suggest that the proposed metrics are effective for measuring distance prediction. We anticipate that this study paves the way for objectively monitoring the progress of inter-residue distance prediction. A web server and a standalone package are provided to implement the proposed metrics. http://yanglab.nankai.edu.cn/APD. Supplementary data are available at Bioinformatics online.