Deep learning the structural determinants of protein biochemical properties by comparing structural ensembles with DiffNets

Privacy-preserving biological age prediction over federated human methylation data using fully homomorphic encryption

DNA methylation data play a crucial role in estimating chronological age in mammals, offering real-time insights into an individual's aging process. The epigenetic pacemaker (EPM) model allows inference of the biological age as deviations from the population trend. Given the sensitivity of this data, it is essential to safeguard both inputs and outputs of the EPM model. A privacy-preserving approach for EPM computation utilizing fully homomorphic encryption was recently introduced. However, this method has limitations, including having high communication complexity and being impractical for large data sets. The current work presents a new privacy-preserving protocol for EPM computation, analytically improving both privacy and complexity. Notably, we employ a single server for the secure computation phase while ensuring privacy even in the event of server corruption (compared to requiring two noncolluding servers in prior work). Using techniques from symbolic algebra and number theory, the new protocol eliminates the need for communication during secure computation, significantly improves asymptotic runtime, and offers better compatibility to parallel computing for further time complexity reduction. We implemented our protocol, demonstrating its ability to produce results similar to the standard (insecure) EPM model with substantial performance improvement compared to prior work. These findings hold promise for enhancing data security in medical applications where personal privacy is paramount. The generality of both the new approach and the EPM suggests that this protocol may be useful in other applications employing similar expectation-maximization techniques.

Complex state transitions of the bacterial cell division protein FtsZ

The key bacterial cell division protein FtsZ can adopt multiple conformations, and prevailing models suggest that transitions of FtsZ subunits from the closed to open state are necessary for filament formation and stability. Using all-atom molecular dynamics simulations, we analyzed state transitions of Staphylococcus aureus FtsZ as a monomer, dimer, and hexamer. We found that monomers can adopt intermediate states but preferentially adopt a closed state that is robust to forced reopening. Dimer subunits transitioned between open and closed states, and dimers with both subunits in the closed state remained highly stable, suggesting that open-state conformations are not necessary for filament formation. Mg2+ strongly stabilized the conformation of GTP-bound subunits and the dimer filament interface. Our hexamer simulations indicate that the plus end subunit preferentially closes and that other subunits can transition between states without affecting inter-subunit stability. We found that rather than being correlated with subunit opening, inter-subunit stability was strongly correlated with catalytic site interactions. By leveraging deep-learning models, we identified key intrasubunit interactions governing state transitions. Our findings suggest a greater range of possible monomer and filament states than previously considered and offer new insights into the nuanced interplay between subunit states and the critical role of nucleotide hydrolysis and Mg2+ in FtsZ filament dynamics.

Sampling Conformational Ensembles of Highly Dynamic Proteins via Generative Deep Learning

Proteins are inherently dynamic, and their conformational ensembles are functionally important in biology. Large-scale motions may govern protein structure-function relationship, and numerous transient but stable conformations of intrinsically disordered proteins (IDPs) can play a crucial role in biological function. Investigating conformational ensembles to understand regulations and disease-related aggregations of IDPs is challenging both experimentally and computationally. In this paper we first introduced an unsupervised deep learning-based model, termed Internal Coordinate Net (ICoN), which learns the physical principles of conformational changes from molecular dynamics (MD) simulation data. Second, we selected interpolating data points in the learned latent space that rapidly identify novel synthetic conformations with sophisticated and large-scale sidechains and backbone arrangements. Third, with the highly dynamic amyloid-β 1-42 (Aβ42) monomer, our deep learning model provided a comprehensive sampling of Aβ42's conformational landscape. Analysis of these synthetic conformations revealed conformational clusters that can be used to rationalize experimental findings. Additionally, the method can identify novel conformations with important interactions in atomistic details that are not included in the training data. New synthetic conformations showed distinct sidechain rearrangements that are probed by our EPR and amino acid substitution studies. This approach is highly transferable and can be used for any available data for training. The work also demonstrated the ability for deep learning to utilize learned natural atomistic motions in protein conformation sampling.

Sampling Conformational Ensembles of Highly Dynamic Proteins via Generative Deep Learning

Proteins are inherently dynamic, and their conformational ensembles are functionally important in biology. Large-scale motions may govern protein structure-function relationship, and numerous transient but stable conformations of intrinsically disordered proteins (IDPs) can play a crucial role in biological function. Investigating conformational ensembles to understand regulations and disease-related aggregations of IDPs is challenging both experimentally and computationally. In this paper first an unsupervised deep learning-based model, termed Internal Coordinate Net (ICoN), is developed that learns the physical principles of conformational changes from molecular dynamics (MD) simulation data. Second, interpolating data points in the learned latent space are selected that rapidly identify novel synthetic conformations with sophisticated and large-scale sidechains and backbone arrangements. Third, with the highly dynamic amyloid-β1-42 (Aβ42) monomer, our deep learning model provided a comprehensive sampling of Aβ42's conformational landscape. Analysis of these synthetic conformations revealed conformational clusters that can be used to rationalize experimental findings. Additionally, the method can identify novel conformations with important interactions in atomistic details that are not included in the training data. New synthetic conformations showed distinct sidechain rearrangements that are probed by our EPR and amino acid substitution studies. The proposed approach is highly transferable and can be used for any available data for training. The work also demonstrated the ability for deep learning to utilize learned natural atomistic motions in protein conformation sampling.

Engineering of the start condensation domain with improved N-decanoyl catalytic activity for daptomycin biosynthesis

Daptomycin, a lipopeptide comprising an N-decanoyl fatty acyl chain and a peptide core, is used clinically as an antimicrobial agent. The start condensation domain (dptC1) is an enzyme that catalyzes the lipoinitiation step of the daptomycin synthesis. In this study, we integrated enzymology, protein engineering, and computer simulation to study the substrate selectivity of the start condensation domain (dptC1) and to screen mutants with improved activity for decanoyl loading. Through molecular docking and computer simulation, the fatty acyl substrate channel and the protein-protein interaction interface of dptC1 are analyzed. Key residues at the protein-protein interface between dptC1 and the acyl carrier were mutated, and a single-point mutant showed more than three-folds improved catalytic efficiency of the target n-decanoyl substrate in comparing with the wild type. Moreover, molecular dynamics simulations suggested that mutants with increased catalytic activity may correlated with a more "open" and contracted substrate binding channel. Our work provides a new perspective for the elucidation of lipopeptide natural products biosynthesis, and also provides new resources to enrich its diversity and optimize the production of important components.

ctGAN: combined transformation of gene expression and survival data with generative adversarial network

Recent studies have extensively used deep learning algorithms to analyze gene expression to predict disease diagnosis, treatment effectiveness, and survival outcomes. Survival analysis studies on diseases with high mortality rates, such as cancer, are indispensable. However, deep learning models are plagued by overfitting owing to the limited sample size relative to the large number of genes. Consequently, the latest style-transfer deep generative models have been implemented to generate gene expression data. However, these models are limited in their applicability for clinical purposes because they generate only transcriptomic data. Therefore, this study proposes ctGAN, which enables the combined transformation of gene expression and survival data using a generative adversarial network (GAN). ctGAN improves survival analysis by augmenting data through style transformations between breast cancer and 11 other cancer types. We evaluated the concordance index (C-index) enhancements compared with previous models to demonstrate its superiority. Performance improvements were observed in nine of the 11 cancer types. Moreover, ctGAN outperformed previous models in seven out of the 11 cancer types, with colon adenocarcinoma (COAD) exhibiting the most significant improvement (median C-index increase of ~15.70%). Furthermore, integrating the generated COAD enhanced the log-rank p-value (0.041) compared with using only the real COAD (p-value = 0.797). Based on the data distribution, we demonstrated that the model generated highly plausible data. In clustering evaluation, ctGAN exhibited the highest performance in most cases (89.62%). These findings suggest that ctGAN can be meaningfully utilized to predict disease progression and select personalized treatments in the medical field.

Interpretation of autoencoder-learned collective variables using Morse-Smale complex and sublevelset persistent homology: An application on molecular trajectories

Dimensionality reduction often serves as the first step toward a minimalist understanding of physical systems as well as the accelerated simulations of them. In particular, neural network-based nonlinear dimensionality reduction methods, such as autoencoders, have shown promising outcomes in uncovering collective variables (CVs). However, the physical meaning of these CVs remains largely elusive. In this work, we constructed a framework that (1) determines the optimal number of CVs needed to capture the essential molecular motions using an ensemble of hierarchical autoencoders and (2) provides topology-based interpretations to the autoencoder-learned CVs with Morse-Smale complex and sublevelset persistent homology. This approach was exemplified using a series of n-alkanes and can be regarded as a general, explainable nonlinear dimensionality reduction method.

Toward physics-based precision medicine: Exploiting protein dynamics to design new therapeutics and interpret variants

The goal of precision medicine is to utilize our knowledge of the molecular causes of disease to better diagnose and treat patients. However, there is a substantial mismatch between the small number of food and drug administration (FDA)-approved drugs and annotated coding variants compared to the needs of precision medicine. This review introduces the concept of physics-based precision medicine, a scalable framework that promises to improve our understanding of sequence-function relationships and accelerate drug discovery. We show that accounting for the ensemble of structures a protein adopts in solution with computer simulations overcomes many of the limitations imposed by assuming a single protein structure. We highlight studies of protein dynamics and recent methods for the analysis of structural ensembles. These studies demonstrate that differences in conformational distributions predict functional differences within protein families and between variants. Thanks to new computational tools that are providing unprecedented access to protein structural ensembles, this insight may enable accurate predictions of variant pathogenicity for entire libraries of variants. We further show that explicitly accounting for protein ensembles, with methods like alchemical free energy calculations or docking to Markov state models, can uncover novel lead compounds. To conclude, we demonstrate that cryptic pockets, or cavities absent in experimental structures, provide an avenue to target proteins that are currently considered undruggable. Taken together, our review provides a roadmap for the field of protein science to accelerate precision medicine.

Homologous mutations in human β, embryonic, and perinatal muscle myosins have divergent effects on molecular power generation

Mutations at a highly conserved homologous residue in three closely related muscle myosins cause three distinct diseases involving muscle defects: R671C in β-cardiac myosin causes hypertrophic cardiomyopathy, R672C and R672H in embryonic skeletal myosin cause Freeman-Sheldon syndrome, and R674Q in perinatal skeletal myosin causes trismus-pseudocamptodactyly syndrome. It is not known whether their effects at the molecular level are similar to one another or correlate with disease phenotype and severity. To this end, we investigated the effects of the homologous mutations on key factors of molecular power production using recombinantly expressed human β, embryonic, and perinatal myosin subfragment-1. We found large effects in the developmental myosins but minimal effects in β myosin, and magnitude of changes correlated partially with clinical severity. The mutations in the developmental myosins dramatically decreased the step size and load-sensitive actin-detachment rate of single molecules measured by optical tweezers, in addition to decreasing overall enzymatic (ATPase) cycle rate. In contrast, the only measured effect of R671C in β myosin was a larger step size. Our measurements of step size and bound times predicted velocities consistent with those measured in an in vitro motility assay. Finally, molecular dynamics simulations predicted that the arginine to cysteine mutation in embryonic, but not β, myosin may reduce pre-powerstroke lever arm priming and ADP pocket opening, providing a possible structural mechanism consistent with the experimental observations. This paper presents direct comparisons of homologous mutations in several different myosin isoforms, whose divergent functional effects are a testament to myosin's highly allosteric nature.

Modulation of Allostery with Multiple Mechanisms by Hotspot Mutations in TetR

Modulating allosteric coupling offers unique opportunities for biomedical applications. Such efforts can benefit from efficient prediction and evaluation of allostery hotspot residues that dictate the degree of cooperativity between distant sites. We demonstrate that effects of allostery hotspot mutations can be evaluated qualitatively and semiquantitatively by molecular dynamics simulations in a bacterial tetracycline repressor (TetR). The simulations recapitulate the effects of these mutations on abolishing the induction function of TetR and provide a rationale for the different rescuabilities observed to restore allosteric coupling of the hotspot mutations. We demonstrate that the same noninducible phenotype could be the result of perturbations in distinct structural and energetic properties of TetR. Our work underscores the value of explicitly computing the functional free energy landscapes to effectively evaluate and rank hotspot mutations despite the prevalence of compensatory interactions and therefore provides quantitative guidance to allostery modulation for therapeutic and engineering applications.

Modulation of Allostery with Multiple Mechanisms by Hotspot Mutations in TetR

Modulating allosteric coupling offers unique opportunities for biomedical applications. Such efforts can benefit from efficient prediction and evaluation of allostery hotspot residues that dictate the degree of co-operativity between distant sites. We demonstrate that effects of allostery hotspot mutations can be evaluated qualitatively and semi-quantitatively by molecular dynamics simulations in a bacterial tetracycline repressor (TetR). The simulations recapitulate the effects of these mutations on abolishing the induction function of TetR and provide a rationale for the different degrees of rescuability observed to restore allosteric coupling of the hotspot mutations. We demonstrate that the same non-inducible phenotype could be the result of perturbations in distinct structural and energetic properties of TetR. Our work underscore the value of explicitly computing the functional free energy landscapes to effectively evaluate and rank hotspot mutations despite the prevalence of compensatory interactions, and therefore provide quantitative guidance to allostery modulation for therapeutic and engineering applications.

From Byte to Bench to Bedside: Molecular Dynamics Simulations and Drug Discovery

Molecular dynamics (MD) simulations and computer-aided drug design (CADD) have advanced substantially over the past two decades, thanks to continuous computer hardware and software improvements. Given these advancements, MD simulations are poised to become even more powerful tools for investigating the dynamic interactions between potential small-molecule drugs and their target proteins, with significant implications for pharmacological research.

Machine Learning-Guided Protein Engineering

Recent progress in engineering highly promising biocatalysts has increasingly involved machine learning methods. These methods leverage existing experimental and simulation data to aid in the discovery and annotation of promising enzymes, as well as in suggesting beneficial mutations for improving known targets. The field of machine learning for protein engineering is gathering steam, driven by recent success stories and notable progress in other areas. It already encompasses ambitious tasks such as understanding and predicting protein structure and function, catalytic efficiency, enantioselectivity, protein dynamics, stability, solubility, aggregation, and more. Nonetheless, the field is still evolving, with many challenges to overcome and questions to address. In this Perspective, we provide an overview of ongoing trends in this domain, highlight recent case studies, and examine the current limitations of machine learning-based methods. We emphasize the crucial importance of thorough experimental validation of emerging models before their use for rational protein design. We present our opinions on the fundamental problems and outline the potential directions for future research.

Discovery of Nonretinoid Inhibitors of CRBP1: Structural and Dynamic Insights for Ligand-Binding Mechanisms

The dysregulation of retinoid metabolism has been linked to prevalent ocular diseases including age-related macular degeneration and Stargardt disease. Modulating retinoid metabolism through pharmacological approaches holds promise for the treatment of these eye diseases. Cellular retinol-binding protein 1 (CRBP1) is the primary transporter of all-trans-retinol (atROL) in the eye, and its inhibition has recently been shown to protect mouse retinas from light-induced retinal damage. In this report, we employed high-throughput screening to identify new chemical scaffolds for competitive, nonretinoid inhibitors of CRBP1. To understand the mechanisms of interaction between CRBP1 and these inhibitors, we solved high-resolution X-ray crystal structures of the protein in complex with six selected compounds. By combining protein crystallography with hydrogen/deuterium exchange mass spectrometry, we quantified the conformational changes in CRBP1 caused by different inhibitors and correlated their magnitude with apparent binding affinities. Furthermore, using molecular dynamic simulations, we provided evidence for the functional significance of the "closed" conformation of CRBP1 in retaining ligands within the binding pocket. Collectively, our study outlines the molecular foundations for understanding the mechanism of high-affinity interactions between small molecules and CRBPs, offering a framework for the rational design of improved inhibitors for this class of lipid-binding proteins.

Exploring Conformational Landscapes and Cryptic Binding Pockets in Distinct Functional States of the SARS-CoV-2 Omicron BA.1 and BA.2 Trimers: Mutation-Induced Modulation of Protein Dynamics and Network-Guided Prediction of Variant-Specific Allosteric Binding Sites

A significant body of experimental structures of SARS-CoV-2 spike trimers for the BA.1 and BA.2 variants revealed a considerable plasticity of the spike protein and the emergence of druggable binding pockets. Understanding the interplay of conformational dynamics changes induced by the Omicron variants and the identification of cryptic dynamic binding pockets in the S protein is of paramount importance as exploring broad-spectrum antiviral agents to combat the emerging variants is imperative. In the current study, we explore conformational landscapes and characterize the universe of binding pockets in multiple open and closed functional spike states of the BA.1 and BA.2 Omicron variants. By using a combination of atomistic simulations, a dynamics network analysis, and an allostery-guided network screening of binding pockets in the conformational ensembles of the BA.1 and BA.2 spike conformations, we identified all experimentally known allosteric sites and discovered significant variant-specific differences in the distribution of binding sites in the BA.1 and BA.2 trimers. This study provided a structural characterization of the predicted cryptic pockets and captured the experimentally known allosteric sites, revealing the critical role of conformational plasticity in modulating the distribution and cross-talk between functional binding sites. We found that mutational and dynamic changes in the BA.1 variant can induce the remodeling and stabilization of a known druggable pocket in the N-terminal domain, while this pocket is drastically altered and may no longer be available for ligand binding in the BA.2 variant. Our results predicted the experimentally known allosteric site in the receptor-binding domain that remains stable and ranks as the most favorable site in the conformational ensembles of the BA.2 variant but could become fragmented and less probable in BA.1 conformations. We also uncovered several cryptic pockets formed at the inter-domain and inter-protomer interface, including functional regions of the S2 subunit and stem helix region, which are consistent with the known role of pocket residues in modulating conformational transitions and antibody recognition. The results of this study are particularly significant for understanding the dynamic and network features of the universe of available binding pockets in spike proteins, as well as the effects of the Omicron-variant-specific modulation of preferential druggable pockets. The exploration of predicted druggable sites can present a new and previously underappreciated opportunity for therapeutic interventions for Omicron variants through the conformation-selective and variant-specific targeting of functional sites involved in allosteric changes.

NCAE: data-driven representations using a deep network-coherent DNA methylation autoencoder identify robust disease and risk factor signatures

Precision medicine relies on the identification of robust disease and risk factor signatures from omics data. However, current knowledge-driven approaches may overlook novel or unexpected phenomena due to the inherent biases in biological knowledge. In this study, we present a data-driven signature discovery workflow for DNA methylation analysis utilizing network-coherent autoencoders (NCAEs) with biologically relevant latent embeddings. First, we explored the architecture space of autoencoders trained on a large-scale pan-tissue compendium (n = 75 272) of human epigenome-wide association studies. We observed the emergence of co-localized patterns in the deep autoencoder latent space representations that corresponded to biological network modules. We determined the NCAE configuration with the strongest co-localization and centrality signals in the human protein interactome. Leveraging the NCAE embeddings, we then trained interpretable deep neural networks for risk factor (aging, smoking) and disease (systemic lupus erythematosus) prediction and classification tasks. Remarkably, our NCAE embedding-based models outperformed existing predictors, revealing novel DNA methylation signatures enriched in gene sets and pathways associated with the studied condition in each case. Our data-driven biomarker discovery workflow provides a generally applicable pipeline to capture relevant risk factor and disease information. By surpassing the limitations of knowledge-driven methods, our approach enhances the understanding of complex epigenetic processes, facilitating the development of more effective diagnostic and therapeutic strategies.

Open-Source Machine Learning in Computational Chemistry

The field of computational chemistry has seen a significant increase in the integration of machine learning concepts and algorithms. In this Perspective, we surveyed 179 open-source software projects, with corresponding peer-reviewed papers published within the last 5 years, to better understand the topics within the field being investigated by machine learning approaches. For each project, we provide a short description, the link to the code, the accompanying license type, and whether the training data and resulting models are made publicly available. Based on those deposited in GitHub repositories, the most popular employed Python libraries are identified. We hope that this survey will serve as a resource to learn about machine learning or specific architectures thereof by identifying accessible codes with accompanying papers on a topic basis. To this end, we also include computational chemistry open-source software for generating training data and fundamental Python libraries for machine learning. Based on our observations and considering the three pillars of collaborative machine learning work, open data, open source (code), and open models, we provide some suggestions to the community.

Folding@home: Achievements from over 20 years of citizen science herald the exascale era

Simulations of biomolecules have enormous potential to inform our understanding of biology but require extremely demanding calculations. For over 20 years, the Folding@home distributed computing project has pioneered a massively parallel approach to biomolecular simulation, harnessing the resources of citizen scientists across the globe. Here, we summarize the scientific and technical advances this perspective has enabled. As the project's name implies, the early years of Folding@home focused on driving advances in our understanding of protein folding by developing statistical methods for capturing long-timescale processes and facilitating insight into complex dynamical processes. Success laid a foundation for broadening the scope of Folding@home to address other functionally relevant conformational changes, such as receptor signaling, enzyme dynamics, and ligand binding. Continued algorithmic advances, hardware developments such as graphics processing unit (GPU)-based computing, and the growing scale of Folding@home have enabled the project to focus on new areas where massively parallel sampling can be impactful. While previous work sought to expand toward larger proteins with slower conformational changes, new work focuses on large-scale comparative studies of different protein sequences and chemical compounds to better understand biology and inform the development of small-molecule drugs. Progress on these fronts enabled the community to pivot quickly in response to the COVID-19 pandemic, expanding to become the world's first exascale computer and deploying this massive resource to provide insight into the inner workings of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus and aid the development of new antivirals. This success provides a glimpse of what is to come as exascale supercomputers come online and as Folding@home continues its work.

Acetylation coordinates the crosstalk between carbon metabolism and ammonium assimilation in Salmonella enterica

Enteric bacteria use up to 15% of their cellular energy for ammonium assimilation via glutamine synthetase (GS)/glutamate synthase (GOGAT) and glutamate dehydrogenase (GDH) in response to varying ammonium availability. However, the sensory mechanisms for effective and appropriate coordination between carbon metabolism and ammonium assimilation have not been fully elucidated. Here, we report that in Salmonella enterica, carbon metabolism coordinates the activities of GS/GDH via functionally reversible protein lysine acetylation. Glucose promotes Pat acetyltransferase-mediated acetylation and activation of adenylylated GS. Simultaneously, glucose induces GDH acetylation to inactivate the enzyme by impeding its catalytic centre, which is reversed upon GDH deacetylation by deacetylase CobB. Molecular dynamics (MD) simulations indicate that adenylylation is required for acetylation-dependent activation of GS. We show that acetylation and deacetylation occur within minutes of "glucose shock" to promptly adapt to ammonium/carbon variation and finely balance glutamine/glutamate synthesis. Finally, in a mouse infection model, reduced S. enterica growth caused by the expression of adenylylation-mimetic GS is rescued by acetylation-mimicking mutations. Thus, glucose-driven acetylation integrates signals from ammonium assimilation and carbon metabolism to fine-tune bacterial growth control.

Homologous mutations in β, embryonic, and perinatal muscle myosins have divergent effects on molecular power generation

Mutations at a highly conserved homologous residue in three closely related muscle myosins cause three distinct diseases involving muscle defects: R671C in β -cardiac myosin causes hypertrophic cardiomyopathy, R672C and R672H in embryonic skeletal myosin cause Freeman Sheldon syndrome, and R674Q in perinatal skeletal myosin causes trismus-pseudocamptodactyly syndrome. It is not known if their effects at the molecular level are similar to one another or correlate with disease phenotype and severity. To this end, we investigated the effects of the homologous mutations on key factors of molecular power production using recombinantly expressed human β , embryonic, and perinatal myosin subfragment-1. We found large effects in the developmental myosins, with the most dramatic in perinatal, but minimal effects in β myosin, and magnitude of changes correlated partially with clinical severity. The mutations in the developmental myosins dramatically decreased the step size and load-sensitive actin-detachment rate of single molecules measured by optical tweezers, in addition to decreasing ATPase cycle rate. In contrast, the only measured effect of R671C in β myosin was a larger step size. Our measurements of step size and bound times predicted velocities consistent with those measured in an in vitro motility assay. Finally, molecular dynamics simulations predicted that the arginine to cysteine mutation in embryonic, but not β , myosin may reduce pre-powerstroke lever arm priming and ADP pocket opening, providing a possible structural mechanism consistent with the experimental observations. This paper presents the first direct comparisons of homologous mutations in several different myosin isoforms, whose divergent functional effects are yet another testament to myosin's highly allosteric nature.

Integrative Approach to Dissect the Drug Resistance Mechanism of the H172Y Mutation of SARS-CoV-2 Main Protease

Nirmatrelvir is an orally available inhibitor of SARS-CoV-2 main protease (Mpro) and the main ingredient of Paxlovid, a drug approved by the U.S. Food and Drug Administration for high-risk COVID-19 patients. Recently, a rare natural mutation, H172Y, was found to significantly reduce nirmatrelvir's inhibitory activity. As the COVID-19 cases skyrocket in China and the selective pressure of antiviral therapy builds in the US, there is an urgent need to characterize and understand how the H172Y mutation confers drug resistance. Here, we investigated the H172Y Mpro's conformational dynamics, folding stability, catalytic efficiency, and inhibitory activity using all-atom constant pH and fixed-charge molecular dynamics simulations, alchemical and empirical free energy calculations, artificial neural networks, and biochemical experiments. Our data suggest that the mutation significantly weakens the S1 pocket interactions with the N-terminus and perturbs the conformation of the oxyanion loop, leading to a decrease in the thermal stability and catalytic efficiency. Importantly, the perturbed S1 pocket dynamics weaken the nirmatrelvir binding in the P1 position, which explains the decreased inhibitory activity of nirmatrelvir. Our work demonstrates the predictive power of the combined simulation and artificial intelligence approaches, and together with biochemical experiments, they can be used to actively surveil continually emerging mutations of SARS-CoV-2 Mpro and assist the optimization of antiviral drugs. The presented approach, in general, can be applied to characterize mutation effects on any protein drug targets.

MDiGest: A Python package for describing allostery from molecular dynamics simulations

Many biological processes are regulated by allosteric mechanisms that communicate with distant sites in the protein responsible for functionality. The binding of a small molecule at an allosteric site typically induces conformational changes that propagate through the protein along allosteric pathways regulating enzymatic activity. Elucidating those communication pathways from allosteric sites to orthosteric sites is, therefore, essential to gain insights into biochemical processes. Targeting the allosteric pathways by mutagenesis can allow the engineering of proteins with desired functions. Furthermore, binding small molecule modulators along the allosteric pathways is a viable approach to target reactions using allosteric inhibitors/activators with temporal and spatial selectivity. Methods based on network theory can elucidate protein communication networks through the analysis of pairwise correlations observed in molecular dynamics (MD) simulations using molecular descriptors that serve as proxies for allosteric information. Typically, single atomic descriptors such as α-carbon displacements are used as proxies for allosteric information. Therefore, allosteric networks are based on correlations revealed by that descriptor. Here, we introduce a Python software package that provides a comprehensive toolkit for studying allostery from MD simulations of biochemical systems. MDiGest offers the ability to describe protein dynamics by combining different approaches, such as correlations of atomic displacements or dihedral angles, as well as a novel approach based on the correlation of Kabsch-Sander electrostatic couplings. MDiGest allows for comparisons of networks and community structures that capture physical information relevant to allostery. Multiple complementary tools for studying essential dynamics include principal component analysis, root mean square fluctuation, as well as secondary structure-based analyses.

From Deep Mutational Mapping of Allosteric Protein Landscapes to Deep Learning of Allostery and Hidden Allosteric Sites: Zooming in on "Allosteric Intersection" of Biochemical and Big Data Approaches

The recent advances in artificial intelligence (AI) and machine learning have driven the design of new expert systems and automated workflows that are able to model complex chemical and biological phenomena. In recent years, machine learning approaches have been developed and actively deployed to facilitate computational and experimental studies of protein dynamics and allosteric mechanisms. In this review, we discuss in detail new developments along two major directions of allosteric research through the lens of data-intensive biochemical approaches and AI-based computational methods. Despite considerable progress in applications of AI methods for protein structure and dynamics studies, the intersection between allosteric regulation, the emerging structural biology technologies and AI approaches remains largely unexplored, calling for the development of AI-augmented integrative structural biology. In this review, we focus on the latest remarkable progress in deep high-throughput mining and comprehensive mapping of allosteric protein landscapes and allosteric regulatory mechanisms as well as on the new developments in AI methods for prediction and characterization of allosteric binding sites on the proteome level. We also discuss new AI-augmented structural biology approaches that expand our knowledge of the universe of protein dynamics and allostery. We conclude with an outlook and highlight the importance of developing an open science infrastructure for machine learning studies of allosteric regulation and validation of computational approaches using integrative studies of allosteric mechanisms. The development of community-accessible tools that uniquely leverage the existing experimental and simulation knowledgebase to enable interrogation of the allosteric functions can provide a much-needed boost to further innovation and integration of experimental and computational technologies empowered by booming AI field.

Resolving Protein Conformational Plasticity and Substrate Binding via Machine Learning

A long-standing target in elucidating the biomolecular recognition process is the identification of binding-competent conformations of the receptor protein. However, protein conformational plasticity and the stochastic nature of the recognition processes often preclude the assignment of a specific protein conformation to an individual ligand-bound pose. Here, we demonstrate that a computational framework coined as RF-TICA-MD, which integrates an ensemble decision-tree-based Random Forest (RF) machine learning (ML) technique with an unsupervised dimension reduction approach time-structured independent component analysis (TICA), provides an efficient and unambiguous solution toward resolving protein conformational plasticity and the substrate binding process. In particular, we consider multimicrosecond-long molecular dynamics (MD) simulation trajectories of a ligand recognition process in solvent-inaccessible cavities of archetypal proteins T4 lysozyme and cytochrome P450cam. We show that in a scenario in which clear correspondence between protein conformation and binding-competent macrostates could not be obtained via an unsupervised dimension reduction approach, an a priori decision-tree-based supervised classification of the simulated recognition trajectories via RF would help characterize key amino acid residue pairs of the protein that are deemed sensitive for ligand binding. A subsequent unsupervised dimensional reduction of the selected residue pairs via TICA would then delineate a conformational landscape of protein which is able to demarcate ligand-bound poses from unbound ones. The proposed RF-TICA-MD approach is shown to be data agnostic and found to be robust when using other ML-based classification methods such as XGBoost. As a promising spinoff of the protocol, the framework is found to be capable of identifying distal protein locations which would be allosterically important for ligand binding and would characterize their roles in recognition pathways. A Python implementation of a proposed ML workflow is available in GitHub https://github.com/navjeet0211/rf-tica-md.

Can molecular dynamics simulations improve predictions of protein-ligand binding affinity with machine learning?

Binding affinity prediction largely determines the discovery efficiency of lead compounds in drug discovery. Recently, machine learning (ML)-based approaches have attracted much attention in hopes of enhancing the predictive performance of traditional physics-based approaches. In this study, we evaluated the impact of structural dynamic information on the binding affinity prediction by comparing the models trained on different dimensional descriptors, using three targets (i.e. JAK1, TAF1-BD2 and DDR1) and their corresponding ligands as the examples. Here, 2D descriptors are traditional ECFP4 fingerprints, 3D descriptors are the energy terms of the Smina and NNscore scoring functions and 4D descriptors contain the structural dynamic information derived from the trajectories based on molecular dynamics (MD) simulations. We systematically investigate the MD-refined binding affinity prediction performance of three classical ML algorithms (i.e. RF, SVR and XGB) as well as two common virtual screening methods, namely Glide docking and MM/PBSA. The outcomes of the ML models built using various dimensional descriptors and their combinations reveal that the MD refinement with the optimized protocol can improve the predictive performance on the TAF1-BD2 target with considerable structural flexibility, but not for the less flexible JAK1 and DDR1 targets, when taking docking poses as the initial structure instead of the crystal structures. The results highlight the importance of the initial structures to the final performance of the model through conformational analysis on the three targets with different flexibility.

Exploring and Learning the Universe of Protein Allostery Using Artificial Intelligence Augmented Biophysical and Computational Approaches

Allosteric mechanisms are commonly employed regulatory tools used by proteins to orchestrate complex biochemical processes and control communications in cells. The quantitative understanding and characterization of allosteric molecular events are among major challenges in modern biology and require integration of innovative computational experimental approaches to obtain atomistic-level knowledge of the allosteric states, interactions, and dynamic conformational landscapes. The growing body of computational and experimental studies empowered by emerging artificial intelligence (AI) technologies has opened up new paradigms for exploring and learning the universe of protein allostery from first principles. In this review we analyze recent developments in high-throughput deep mutational scanning of allosteric protein functions; applications and latest adaptations of Alpha-fold structural prediction methods for studies of protein dynamics and allostery; new frontiers in integrating machine learning and enhanced sampling techniques for characterization of allostery; and recent advances in structural biology approaches for studies of allosteric systems. We also highlight recent computational and experimental studies of the SARS-CoV-2 spike (S) proteins revealing an important and often hidden role of allosteric regulation driving functional conformational changes, binding interactions with the host receptor, and mutational escape mechanisms of S proteins which are critical for viral infection. We conclude with a summary and outlook of future directions suggesting that AI-augmented biophysical and computer simulation approaches are beginning to transform studies of protein allostery toward systematic characterization of allosteric landscapes, hidden allosteric states, and mechanisms which may bring about a new revolution in molecular biology and drug discovery.

Characterization of RNA polymerase II trigger loop mutations using molecular dynamics simulations and machine learning

Catalysis and fidelity of multisubunit RNA polymerases rely on a highly conserved active site domain called the trigger loop (TL), which achieves roles in transcription through conformational changes and interaction with NTP substrates. The mutations of TL residues cause distinct effects on catalysis including hypo- and hyperactivity and altered fidelity. We applied molecular dynamics simulation (MD) and machine learning (ML) techniques to characterize TL mutations in the Saccharomyces cerevisiae RNA Polymerase II (Pol II) system. We did so to determine relationships between individual mutations and phenotypes and to associate phenotypes with MD simulated structural alterations. Using fitness values of mutants under various stress conditions, we modeled phenotypes along a spectrum of continual values. We found that ML could predict the phenotypes with 0.68 R2 correlation from amino acid sequences alone. It was more difficult to incorporate MD data to improve predictions from machine learning, presumably because MD data is too noisy and possibly incomplete to directly infer functional phenotypes. However, a variational auto-encoder model based on the MD data allowed the clustering of mutants with different phenotypes based on structural details. Overall, we found that a subset of loss-of-function (LOF) and lethal mutations tended to increase distances of TL residues to the NTP substrate, while another subset of LOF and lethal substitutions tended to confer an increase in distances between TL and bridge helix (BH). In contrast, some of the gain-of-function (GOF) mutants appear to cause disruption of hydrophobic contacts among TL and nearby helices.

MEN1 mutations mediate clinical resistance to menin inhibition

Chromatin-binding proteins are critical regulators of cell state in haematopoiesis1,2. Acute leukaemias driven by rearrangement of the mixed lineage leukaemia 1 gene (KMT2Ar) or mutation of the nucleophosmin gene (NPM1) require the chromatin adapter protein menin, encoded by the MEN1 gene, to sustain aberrant leukaemogenic gene expression programs3-5. In a phase 1 first-in-human clinical trial, the menin inhibitor revumenib, which is designed to disrupt the menin-MLL1 interaction, induced clinical responses in patients with leukaemia with KMT2Ar or mutated NPM1 (ref. 6). Here we identified somatic mutations in MEN1 at the revumenib-menin interface in patients with acquired resistance to menin inhibition. Consistent with the genetic data in patients, inhibitor-menin interface mutations represent a conserved mechanism of therapeutic resistance in xenograft models and in an unbiased base-editor screen. These mutants attenuate drug-target binding by generating structural perturbations that impact small-molecule binding but not the interaction with the natural ligand MLL1, and prevent inhibitor-induced eviction of menin and MLL1 from chromatin. To our knowledge, this study is the first to demonstrate that a chromatin-targeting therapeutic drug exerts sufficient selection pressure in patients to drive the evolution of escape mutants that lead to sustained chromatin occupancy, suggesting a common mechanism of therapeutic resistance.

Mutually beneficial confluence of structure-based modeling of protein dynamics and machine learning methods

Proteins sample an ensemble of conformers under physiological conditions, having access to a spectrum of modes of motions, also called intrinsic dynamics. These motions ensure the adaptation to various interactions in the cell, and largely assist in, if not determine, viable mechanisms of biological function. In recent years, machine learning frameworks have proven uniquely useful in structural biology, and recent studies further provide evidence to the utility and/or necessity of considering intrinsic dynamics for increasing their predictive ability. Efficient quantification of dynamics-based attributes by recently developed physics-based theories and models such as elastic network models provides a unique opportunity to generate data on dynamics for training ML models towards inferring mechanisms of protein function, assessing pathogenicity, or estimating binding affinities.

Functional divergence of the sarcomeric myosin, MYH7b, supports species-specific biological roles

Myosin heavy chain 7b (MYH7b) is an evolutionarily ancient member of the sarcomeric myosin family, which typically supports striated muscle function. However, in mammals, alternative splicing prevents MYH7b protein production in cardiac and most skeletal muscles and limits expression to a subset of specialized muscles and certain nonmuscle environments. In contrast, MYH7b protein is abundant in python cardiac and skeletal muscles. Although the MYH7b expression pattern diverges in mammals versus reptiles, MYH7b shares high sequence identity across species. So, it remains unclear how mammalian MYH7b function may differ from that of other sarcomeric myosins and whether human and python MYH7b motor functions diverge as their expression patterns suggest. Thus, we generated recombinant human and python MYH7b protein and measured their motor properties to investigate any species-specific differences in activity. Our results reveal that despite having similar working strokes, the MYH7b isoforms have slower actin-activated ATPase cycles and actin sliding velocities than human cardiac β-MyHC. Furthermore, python MYH7b is tuned to have slower motor activity than human MYH7b because of slower kinetics of the chemomechanical cycle. We found that the MYH7b isoforms adopt a higher proportion of myosin heads in the ultraslow, super-relaxed state compared with human cardiac β-MyHC. These findings are supported by molecular dynamics simulations that predict MYH7b preferentially occupies myosin active site conformations similar to those observed in the structurally inactive state. Together, these results suggest that MYH7b is specialized for slow and energy-conserving motor activity and that differential tuning of MYH7b orthologs contributes to species-specific biological roles.

Multiple conserved states characterize the twist landscape of the bacterial actin homolog MreB

Filament formation by cytoskeletal proteins is critical to their involvement in myriad cellular processes. The bacterial actin homolog MreB, which is essential for cell-shape determination in many rod-shaped bacteria, has served as a model system for studying the mechanics of cytoskeletal filaments. Previous molecular dynamics (MD) simulations revealed that the twist of MreB double protofilaments is dependent on the bound nucleotide, as well as binding to the membrane or the accessory protein RodZ, and MreB mutations that modulate twist also affect MreB spatial organization and cell shape. Here, we show that MreB double protofilaments can adopt multiple twist states during microsecond-scale MD simulations. A deep learning algorithm trained only on high- and low-twist states robustly identified all twist conformations across most perturbations of ATP-bound MreB, suggesting the existence of a conserved set of states whose occupancy is affected by each perturbation to MreB. Simulations replacing ATP with ADP indicated that twist states were generally stable after hydrolysis. These findings suggest a rich twist landscape that could provide the capacity to tune MreB activity and therefore its effects on cell shape.

Recent Advances in Machine Learning Variant Effect Prediction Tools for Protein Engineering

Proteins are Nature's molecular machinery and comprise diverse roles while consisting of chemically similar building blocks. In recent years, protein engineering and design have become important research areas, with many applications in the pharmaceutical, energy, and biocatalysis fields, among others-where the aim is to ultimately create a protein given desired structural and functional properties. It is often critical to model the relationship between a protein's sequence, folded structure, and biological function to assist in such protein engineering pursuits. However, significant challenges remain in concretely mapping an amino acid sequence to specific protein properties and biological activities. Mutations may enhance or diminish molecular protein function, and the epistatic interactions between mutations result in an inherently complex mapping between genetic modifications and protein function. Therefore, estimating the quantitative effects of mutations on protein function(s) remains a grand challenge of biology, bioinformatics, and many related fields and would rapidly accelerate protein engineering tasks when successful. Such estimation is often known as variant effect prediction (VEP). However, progress has been demonstrated in recent years with the development of machine learning (ML) methods in modeling the relationship between mutations and protein function. In this Review, recent advances in variant effect prediction (VEP) are discussed as tools for protein engineering, focusing on techniques incorporating gains from the broader ML community and challenges in estimating biomolecular functional differences. Primary developments highlighted include convolutional neural networks, graph neural networks, and natural language embeddings for protein sequences.

A cryptic pocket in Ebola VP35 allosterically controls RNA binding

Protein-protein and protein-nucleic acid interactions are often considered difficult drug targets because the surfaces involved lack obvious druggable pockets. Cryptic pockets could present opportunities for targeting these interactions, but identifying and exploiting these pockets remains challenging. Here, we apply a general pipeline for identifying cryptic pockets to the interferon inhibitory domain (IID) of Ebola virus viral protein 35 (VP35). VP35 plays multiple essential roles in Ebola's replication cycle but lacks pockets that present obvious utility for drug design. Using adaptive sampling simulations and machine learning algorithms, we predict VP35 harbors a cryptic pocket that is allosterically coupled to a key dsRNA-binding interface. Thiol labeling experiments corroborate the predicted pocket and mutating the predicted allosteric network supports our model of allostery. Finally, covalent modifications that mimic drug binding allosterically disrupt dsRNA binding that is essential for immune evasion. Based on these results, we expect this pipeline will be applicable to other proteins.

Local Ion Densities can Influence Transition Paths of Molecular Binding

Improper reaction coordinates can pose significant problems for path-based binding free energy calculations. Particularly, omission of long timescale motions can lead to over-estimation of the energetic barriers between the bound and unbound states. Many methods exist to construct the optimal reaction coordinate using a pre-defined basis set of features. Although simulations are typically conducted in explicit solvent, the solvent atoms are often excluded by these feature sets—resulting in little being known about their role in reaction coordinates, and ultimately, their role in determining (un)binding rates and free energies. In this work, analysis is done on an extensive set of host-guest unbinding trajectories, working to characterize differences between high and low probability unbinding trajectories with a focus on solvent-based features, including host-ion interactions, guest-ion interactions and location-dependent ion densities. We find that differences in ion densities as well as guest-ion interactions strongly correlate with differences in the probabilities of reactive paths that are used to determine free energies of (un)binding and play a significant role in the unbinding process.

RPnet: a reverse-projection-based neural network for coarse-graining metastable conformational states for protein dynamics

The Markov State Model (MSM) is a powerful tool for modeling long timescale dynamics based on numerous short molecular dynamics (MD) simulation trajectories, which makes it a useful tool for elucidating the conformational changes of biological macromolecules. By partitioning the phase space into discretized states and estimating the probabilities of inter-state transitions based on short MD trajectories, one can construct a kinetic network model that could be used to extrapolate long-timescale kinetics if the Markovian condition is met. However, meeting the Markovian condition often requires hundreds or even thousands of states (microstates), which greatly hinders the comprehension of the conformational dynamics of complex biomolecules. Kinetic lumping algorithms can coarse grain numerous microstates into a handful of metastable states (macrostates), which would greatly facilitate the elucidation of biological mechanisms. In this work, we have developed a reverse-projection-based neural network (RPnet) to lump microstates into macrostates, by making use of a physics-based loss function that is based on the projection operator framework of conformational dynamics. By recognizing that microstate and macrostate transition modes can be related through a projection process, we have developed a reverse-projection scheme to directly compare the microstate and macrostate dynamics. Based on this reverse-projection scheme, we designed a loss function that allows the effective assessment of the quality of a given kinetic lumping. We then make use of a neural network to efficiently minimize this loss function to obtain an optimized set of macrostates. We have demonstrated the power of our RPnet in analyzing the dynamics of a numerical 2D potential, alanine dipeptide, and the clamp opening of an RNA polymerase. In all these systems, we have illustrated that our method could yield comparable or better results than competing methods in terms of state partitioning and reproduction of slow dynamics. We expect that our RPnet holds promise in analyzing the conformational dynamics of biological macromolecules.

Progress in deep Markov state modeling: Coarse graining and experimental data restraints

Recent advances in deep learning frameworks have established valuable tools for analyzing the long-timescale behavior of complex systems, such as proteins. In particular, the inclusion of physical constraints, e.g., time-reversibility, was a crucial step to make the methods applicable to biophysical systems. Furthermore, we advance the method by incorporating experimental observables into the model estimation showing that biases in simulation data can be compensated for. We further develop a new neural network layer in order to build a hierarchical model allowing for different levels of details to be studied. Finally, we propose an attention mechanism, which highlights important residues for the classification into different states. We demonstrate the new methodology on an ultralong molecular dynamics simulation of the Villin headpiece miniprotein.

Variational embedding of protein folding simulations using Gaussian mixture variational autoencoders

Conformational sampling of biomolecules using molecular dynamics simulations often produces a large amount of high dimensional data that makes it difficult to interpret using conventional analysis techniques. Dimensionality reduction methods are thus required to extract useful and relevant information. Here, we devise a machine learning method, Gaussian mixture variational autoencoder (GMVAE), that can simultaneously perform dimensionality reduction and clustering of biomolecular conformations in an unsupervised way. We show that GMVAE can learn a reduced representation of the free energy landscape of protein folding with highly separated clusters that correspond to the metastable states during folding. Since GMVAE uses a mixture of Gaussians as its prior, it can directly acknowledge the multi-basin nature of the protein folding free energy landscape. To make the model end-to-end differentiable, we use a Gumbel-softmax distribution. We test the model on three long-timescale protein folding trajectories and show that GMVAE embedding resembles the folding funnel with folded states down the funnel and unfolded states outside the funnel path. Additionally, we show that the latent space of GMVAE can be used for kinetic analysis and Markov state models built on this embedding produce folding and unfolding timescales that are in close agreement with other rigorous dynamical embeddings such as time independent component analysis.

Naturally Occurring Genetic Variants in the Oxytocin Receptor Alter Receptor Signaling Profiles

The hormone oxytocin is commonly administered during childbirth to initiate and strengthen uterine contractions and prevent postpartum hemorrhage. However, patients have wide variation in the oxytocin dose required for a clinical response. To begin to uncover the mechanisms underlying this variability, we screened the 11 most prevalent missense genetic variants in the oxytocin receptor (OXTR) gene. We found that five variants, V45L, P108A, L206V, V281M, and E339K, significantly altered oxytocin-induced Ca²⁺ signaling or β-arrestin recruitment and proceeded to assess the effects of these variants on OXTR trafficking to the cell membrane, desensitization, and internalization. The variants P108A and L206V increased OXTR localization to the cell membrane, whereas V281M and E339K caused OXTR to be retained inside the cell. We examined how the variants altered the balance between OXTR activation and desensitization, which is critical for appropriate oxytocin dosing. The E339K variant impaired OXTR activation, internalization, and desensitization to roughly equal extents. In contrast, V281M decreased OXTR activation but had no effect on internalization and desensitization. V45L and P108A did not alter OXTR activation but did impair β-arrestin recruitment, internalization, and desensitization. Molecular dynamics simulations predicted that V45L and P108A prevent extension of the first intracellular loop of OXTR, thus inhibiting β-arrestin binding. Overall, our data suggest mechanisms by which OXTR genetic variants could alter clinical response to oxytocin.

Markov state modeling of membrane transport proteins

The flux of ions and molecules in and out of the cell is vital for maintaining the basis of various biological processes. The permeation of substrates across the cellular membrane is mediated through the function of specialized integral membrane proteins commonly known as membrane transporters. These proteins undergo a series of structural rearrangements that allow a primary substrate binding site to be accessed from either side of the membrane at a given time. Structural insights provided by experimentally resolved structures of membrane transporters have aided in the biophysical characterization of these important molecular drug targets. However, characterizing the transitions between conformational states remains challenging to achieve both experimentally and computationally. Though molecular dynamics simulations are a powerful approach to provide atomistic resolution of protein dynamics, a recurring challenge is its ability to efficiently obtain relevant timescales of large conformational transitions as exhibited in transporters. One approach to overcome this difficulty is to adaptively guide the simulation to favor exploration of the conformational landscape, otherwise known as adaptive sampling. Furthermore, such sampling is greatly benefited by the statistical analysis of Markov state models. Historically, the use of Markov state models has been effective in quantifying slow dynamics or long timescale behaviors such as protein folding. Here, we review recent implementations of adaptive sampling and Markov state models to not only address current limitations of molecular dynamics simulations, but to also highlight how Markov state modeling can be applied to investigate the structure-function mechanisms of large, complex membrane transporters.