Molecular Insights from Conformational Ensembles via Machine Learning

A beginner's approach to deep learning applied to VS and MD techniques

It has become impossible to imagine the fields of biochemistry and medicinal chemistry without computational chemistry and molecular modelling techniques. In many steps of the drug development process in silico methods have become indispensable. Virtual screening (VS) can tremendously expedite the early discovery phase, whilst the use of molecular dynamics (MD) simulations forms a powerful additional tool to in vitro methods throughout the entire drug discovery process. In the field of biochemistry, MD has also become a compelling method for studying biophysical systems (e.g., protein folding) complementary to experimental techniques. However, both VS and MD come with their own limitations and methodological difficulties, from hardware limitations to restrictions in algorithmic capabilities. One solution to overcoming these difficulties lies in the field of machine learning (ML), and more specifically deep learning (DL). There are many ways in which DL can be applied to these molecular modelling techniques to achieve more accurate results in a more efficient manner or expedite the data analysis of the acquired results. Despite steadily increasing interest in DL amidst computational chemists, knowledge is still limited and scattered over different resources. This review is aimed at computational chemists with knowledge of molecular modelling, who wish to possibly integrate DL approaches in their research and already have a basic understanding of the fundamentals of DL. This review focusses on a survey of recent applications of DL in molecular modelling techniques. The different sections are logically subdivided, based on where DL is integrated in the research: (1) for the improvement of VS workflows, (2) for the improvement of certain workflows in MD simulations, (3) for aiding in the calculations of interatomic forces, or (4) for data analysis of MD trajectories. It will become clear that DL has the capacity to completely transform the way molecular modelling is carried out.

Machine learning of molecular dynamics simulations provides insights into modulation of viral capsid assembly

An effective approach in the development of novel antivirals is to target the assembly of viral capsids using capsid assembly modulators (CAMs). CAMs targeting hepatitis B virus (HBV) have two major modes of function: they can either accelerate nucleocapsid assembly, retaining its structure, or misdirect it into non-capsid-like particles. Previous molecular dynamics (MD) simulations of early capsid-assembly intermediates showed differences in protein conformations for apo and bound states. Here, we have developed and tested several classification machine learning (ML) models to better distinguish between apo-tetramer intermediates and those bound to accelerating or misdirecting CAMs. Models based on tertiary structural properties of the Cp149 tetramers and their inter-dimer orientation, as well as models based on direct and inverse contact distances between protein residues, were tested. All models distinguished the apo states and the two CAM-bound states with high accuracy. Furthermore, tertiary structure models and residue-distance models highlighted different tetramer regions as important for classification. Both models can be used to better understand structural transitions that govern the assembly of nucleocapsids and to assist the development of more potent CAMs. Finally, we demonstrate the utility of classification ML methods in comparing MD trajectories and describe our ML approaches, which can be extended to other systems of interest.

Molecular dynamics and machine learning stratify motion-dependent activity profiles of S-layer destabilizing nanobodies

Nanobody (Nb)-induced disassembly of surface array protein (Sap) S-layers, a two-dimensional paracrystalline protein lattice from Bacillus anthracis, has been presented as a therapeutic intervention for lethal anthrax infections. However, only a subset of existing Nbs with affinity to Sap exhibit depolymerization activity, suggesting that affinity and epitope recognition are not enough to explain inhibitory activity. In this study, we performed all-atom molecular dynamics simulations of each Nb bound to the Sap binding site and trained a collection of machine learning classifiers to predict whether each Nb induces depolymerization. We used feature importance analysis to filter out unnecessary features and engineered remaining features to regularize the feature landscape and encourage learning of the depolymerization mechanism. We find that, while not enforced in training, a gradient-boosting decision tree is able to reproduce the experimental activities of inhibitory Nbs while maintaining high classification accuracy, whereas neural networks were only able to discriminate between classes. Further feature analysis revealed that inhibitory Nbs restrain Sap motions toward an inhibitory conformational state described by domain-domain clamping and induced twisting of domains normal to the lattice plane. We believe these motions drive Sap lattice depolymerization and can be used as design targets for improved Sap-inhibitory Nbs. Finally, we expect our method of study to apply to S-layers that serve as virulence factors in other pathogens, paving the way forward for Nb therapeutics that target depolymerization mechanisms.

Mechanistic insights into P-glycoprotein ligand transport and inhibition revealed by enhanced molecular dynamics simulations

P-glycoprotein (P-gp) plays a crucial role in cellular detoxification and drug efflux processes, transitioning between inward-facing (IF) open, occluded, and outward-facing (OF) states to facilitate substrate transport. Its role is critical in cancer therapy, where P-gp contributes to the multidrug resistance phenotype. In our study, classical and enhanced molecular dynamics (MD) simulations were conducted to dissect the structural and functional features of the P-gp conformational states. Our advanced MD simulations, including kinetically excited targeted MD (ketMD) and adiabatic biasing MD (ABMD), provided deeper insights into state transition and translocation mechanisms. Our findings suggest that the unkinking of TM4 and TM10 helices is a prerequisite for correctly achieving the outward conformation. Simulations of the IF-occluded conformations, characterized by kinked TM4 and TM10 helices, consistently demonstrated altered communication between the transmembrane domains (TMDs) and nucleotide binding domain 2 (NBD2), suggesting the implication of this interface in inhibiting P-gp's efflux function. A particular emphasis was placed on the unstructured linker segment connecting the NBD1 to TMD2 and its role in the transporter's dynamics. With the linker present, we specifically noticed a potential entrance of cholesterol (CHOL) through the TM4-TM6 portal, shedding light on crucial residues involved in accommodating CHOL. We therefore suggest that this entry mechanism could be employed for some P-gp substrates or inhibitors. Our results provide critical data for understanding P-gp functioning and developing new P-gp inhibitors for establishing more effective strategies against multidrug resistance.

A Deep Learning Approach to Uncover Voltage-Gated Ion Channels' Intermediate States

Owing to recent advancements in cryo-electron microscopy, voltage-gated ion channels have gained a greater comprehension of their structural characteristics. However, a significant enigma remains unsolved for a large majority of these channels: their gating mechanism. This mechanism, which encompasses the conformational changes between open and closed states, is pivotal to their proper functioning. Beyond the binary states of open and closed, an ensemble of intermediate states defines the transition path in-between. Due to the lack of experimental data, one might resort to molecular dynamics simulations as an alternative to decipher these states and the transitions between them. However, the high-energy barriers and the colossal time scales involved hinder access to the latter. We present here an application of deep learning as a reliable pipeline for a comprehensive exploration of voltage-gated ion channel conformational rearrangements during gating. We showcase the pipeline performance specifically on the Kv1.2 voltage sensor domain and confront the results with existing data. We demonstrate how our physics-based deep learning approach contributes to the theoretical understanding of these channels and how it might provide further insights into the exploration of channelopathies.

Molecular Dynamics Reveals Altered Interactions between Belzutifan and HIF-2 with Natural Variant G323E or Proximal Phosphorylation at T324

In patients with von-Hippel Lindau (VHL) disease, hypoxia-independent accumulation of HIF-2α leads to increased transcriptional activity of HIF-2α:ARNT that drives cancers such as renal cell carcinoma. Belzutifan, a recently FDA-approved drug, is designed to prevent the transcriptional activity of HIF-2α:ARNT, thereby overcoming the consequences of its unnatural accumulation in VHL-dependent cancers. Emerging evidence suggests that the naturally occurring variant G323E located in the HIF-2α drug binding pocket prevents inhibitory activity of belzutifan analogs, though the mechanism of inhibition remains unclear. Interestingly, proximal phosphorylation at neighboring T324, previously shown to regulate HIF-2 protein interactions, has also been proposed to affect HIF-2 drug binding. Here, we used molecular dynamics (MD) simulations to understand and compare the molecular-level effects of G323E and phospho-T324 (pT324) on the belzutifan bound-HIF-2α:ARNT complex. We find that both G323E and pT324 increase structural flexibility within the drug binding site and reduce the apparent binding affinity for belzutifan. Whereas the effects of G323E are concentrated in the binding pocket Fα helix within the HIF-2α PAS-B domain, pT324 decreased the belzutifan binding affinity and stabilized the HIF-2 heterodimer through an alternate mechanism involving polar interactions between the HIF-2α PAS-B and PAS-A domains. Further analysis via ensemble machine learning uncovered important and distinct interchain residue interactions modified by G323E and pT324. These findings reveal a molecular mechanism of G323E-induced drug resistance and suggest that pT324 may also affect the efficacy of HIF-2 drug binding interactions via allosteric effects.

Thermodynamics-inspired explanations of artificial intelligence

In recent years, predictive machine learning models have gained prominence across various scientific domains. However, their black-box nature necessitates establishing trust in them before accepting their predictions as accurate. One promising strategy involves employing explanation techniques that elucidate the rationale behind a model's predictions in a way that humans can understand. However, assessing the degree of human interpretability of these explanations is a nontrivial challenge. In this work, we introduce interpretation entropy as a universal solution for evaluating the human interpretability of any linear model. Using this concept and drawing inspiration from classical thermodynamics, we present Thermodynamics-inspired Explainable Representations of AI and other black-box Paradigms, a method for generating optimally human-interpretable explanations in a model-agnostic manner. We demonstrate the wide-ranging applicability of this method by explaining predictions from various black-box model architectures across diverse domains, including molecular simulations, text, and image classification.

Adaptive Workflows of Machine Learning Illuminate the Sequential Operation Mechanism of the TAK1's Allosteric Network

Allostery is a fundamental mechanism driving biomolecular processes that holds significant therapeutic concern. Our study rigorously investigates how two distinct machine-learning algorithms uniquely classify two already close-to-active DFG-in states of TAK1, differing just by the presence or absence of its allosteric activator TAB1, from an ensemble mixture of conformations (obtained from 2.4 μs molecular dynamics (MD) simulations). The novelty, however, lies in understanding the deeper algorithmic potentials to systematically derive a diverse set of differential residue connectivity features that reconstruct the essential mechanistic architecture for TAK1-TAB1 allostery in such a close-to-active biochemical scenario. While the recursive, random forest-based workflow displays the potential of conducting discretized, hierarchical derivation of allosteric features, a multilayer perceptron-based approach gains considerable efficacy in revealing fluid connected patterns of features when hybridized with mutual information scoring. Interestingly, both pipelines benchmark similar directions of functional conformational changes for TAK1's activation. The findings significantly advance the depth of mechanistic understanding by highlighting crucial activation signatures along a directed C-lobe → activation loop → ATP pocket channel of information flow, including (1) the αF-αE biterminal alignments and (2) the "catalytic" drift of the activation loop toward kinase active site. Besides, some novel allosteric hotspots (K253, Y206, N189, etc.) are further recognized as TAB1 sensors, transducers, and responders, including a benchmark E70 mutation site, precisely mapping the important structural segments for sequential allosteric execution. Hence, our work demonstrates how to navigate through greater structural depths and dimensions of dynamic allosteric machineries just by leveraging standard ML methods in suitable streamlined workflows adaptive to the specific system and objectives.

Exploring biased activation characteristics by molecular dynamics simulation and machine learning for the μ-opioid receptor

Biased ligands selectively activating specific downstream signaling pathways (termed as biased activation) exhibit significant therapeutic potential. However, the conformational characteristics revealed are very limited for the biased activation, which is not conducive to biased drug development. Motivated by the issue, we combine extensive accelerated molecular dynamics simulations and an interpretable deep learning model to probe the biased activation features for two complex systems constructed by the inactive μOR and two different biased agonists (G-protein-biased agonist TRV130 and β-arrestin-biased agonist endomorphin2). The results indicate that TRV130 binds deeper into the receptor core compared to endomorphin2, located between W2936.48 and D1142.50, and forms hydrogen bonding with D1142.50, while endomorphin2 binds above W2936.48. The G protein-biased agonist induces greater outward movements of the TM6 intracellular end, forming a typical active conformation, while the β-arrestin-biased agonist leads to a smaller extent of outward movements of TM6. Compared with TRV130, endomorphin2 causes more pronounced inward movements of the TM7 intracellular end and more complex conformational changes of H8 and ICL1. In addition, important residues determining the two different biased activation states were further identified by using an interpretable deep learning classification model, including some common biased activation residues across Class A GPCRs like some key residues on the TM2 extracellular end, ECL2, TM5 intracellular end, TM6 intracellular end, and TM7 intracellular end, and some specific important residues of ICL3 for μOR. The observations will provide valuable information for understanding the biased activation mechanism for GPCRs.

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.

Binding to nucleosome poises human SIRT6 for histone H3 deacetylation

Sirtuin 6 (SIRT6) is an NAD+-dependent histone H3 deacetylase that is prominently found associated with chromatin, attenuates transcriptionally active promoters and regulates DNA repair, metabolic homeostasis and lifespan. Unlike other sirtuins, it has low affinity to free histone tails but demonstrates strong binding to nucleosomes. It is poorly understood how SIRT6 docking on nucleosomes stimulates its histone deacetylation activity. Here, we present the structure of human SIRT6 bound to a nucleosome determined by cryogenic electron microscopy. The zinc finger domain of SIRT6 associates tightly with the acidic patch of the nucleosome through multiple arginine anchors. The Rossmann fold domain binds to the terminus of the looser DNA half of the nucleosome, detaching two turns of the DNA from the histone octamer and placing the NAD+ binding pocket close to the DNA exit site. This domain shows flexibility with respect to the fixed zinc finger and moves with, but also relative to, the unwrapped DNA terminus. We apply molecular dynamics simulations of the histone tails in the nucleosome to show that in this mode of interaction, the active site of SIRT6 is perfectly poised to catalyze deacetylation of the H3 histone tail and that the partial unwrapping of the DNA allows even lysines close to the H3 core to reach the enzyme.

GPCR molecular dynamics forecasting using recurrent neural networks

G protein-coupled receptors (GPCRs) are a large superfamily of cell membrane proteins that play an important physiological role as transmitters of extracellular signals. Signal transmission through the cell membrane depends on conformational changes in the transmembrane region of the receptor, which makes the investigation of the dynamics in these regions particularly relevant. Molecular dynamics (MD) simulations provide a wealth of data about the structure, dynamics, and physiological function of biological macromolecules by modelling the interactions between their atomic constituents. In this study, a Recurrent and Convolutional Neural Network (RNN) model, namely Long Short-Term Memory (LSTM), is used to predict the dynamics of two GPCR states and three specific simulations of each one, through their activation path and focussing on specific receptor regions. Active and inactive states of the GPCRs are analysed in six scenarios involving APO, Full Agonist (BI 167107) and Partial Inverse Agonist (carazolol) of the receptor. Four Machine Learning models with increasing complexity in terms of neural network architecture are evaluated, and their results discussed. The best method achieves an overall RMSD lower than 0.139 Å and the transmembrane helices are the regions showing the minimum prediction errors and minimum relative movements of the protein.

Toward a Benchmark for Markov State Models: The Folding of HP35

Adopting a 300 μs long MD trajectory of the folding of villin headpiece (HP35) by D. E. Shaw Research, we recently constructed a Markov state model (MSM) based on inter-residue contacts. The model reproduces the folding time and predicts that the native basin and unfolded region consist of metastable substates that are structurally well-characterized. Recognizing the need to establish well-defined benchmark problems, we study to what extent and in what sense this MSM can be employed as a reference model. Hence, we test the robustness of the MSM by comparing it to models that use alternative combinations of features, dimensionality reduction methods, and clustering schemes. The study suggests some main characteristics of the folding of HP35 that should be reproduced by other competitive models. Moreover, the discussion reveals which parts of the MSM workflow matter most for the considered problem and illustrates the promises and pitfalls of state-based models for the interpretation of biomolecular simulations.

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.

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.

Revealing the Molecular Interactions between Human ACE2 and the Receptor Binding Domain of the SARS-CoV-2 Wild-Type, Alpha and Delta Variants

After a sudden and first spread of the pandemic caused by the novel SARS-CoV-2 (Severe Acute Respiratory Syndrome-Coronavirus 2) wild-type strain, mutants have emerged which have been associated with increased infectivity, inducing surges in the contagions. The first of the so-called variants of concerns, was firstly isolated in the United Kingdom and later renamed Alpha variant. Afterwards, in the middle of 2021, a new variant appeared called Delta. The latter is characterized by the presence of point mutations in the Spike protein of SARS-CoV-2, especially in the Receptor Binding Domain (RBD). When in its active conformation, the RBD can interact with the human receptor Angiotensin-Converting Enzyme 2 (ACE2) to allow the entry of the virions into cells. In this contribution, by using extended all-atom molecular dynamic simulations, complemented with machine learning post-processing, we analyze the changes in the molecular interaction network induced by these different strains in comparison with the wild-type. On one hand, although relevant variations are evidenced, only limited changes in the global stability indicators and in the flexibility profiles have been observed. On the other hand, key differences were obtained by tracking hydrophilic and hydrophobic molecular interactions, concerning both positioning at the ACE2/RBD interface and formation/disruption dynamic behavior.

Markov state modelling reveals heterogeneous drug-inhibition mechanism of Calmodulin

Calmodulin (CaM) is a calcium sensor which binds and regulates a wide range of target-proteins. This implicitly enables the concentration of calcium to influence many downstream physiological responses, including muscle contraction, learning and depression. The antipsychotic drug trifluoperazine (TFP) is a known CaM inhibitor. By binding to various sites, TFP prevents CaM from associating to target-proteins. However, the molecular and state-dependent mechanisms behind CaM inhibition by drugs such as TFP are largely unknown. Here, we build a Markov state model (MSM) from adaptively sampled molecular dynamics simulations and reveal the structural and dynamical features behind the inhibitory mechanism of TFP-binding to the C-terminal domain of CaM. We specifically identify three major TFP binding-modes from the MSM macrostates, and distinguish their effect on CaM conformation by using a systematic analysis protocol based on biophysical descriptors and tools from machine learning. The results show that depending on the binding orientation, TFP effectively stabilizes features of the calcium-unbound CaM, either affecting the CaM hydrophobic binding pocket, the calcium binding sites or the secondary structure content in the bound domain. The conclusions drawn from this work may in the future serve to formulate a complete model of pharmacological modulation of CaM, which furthers our understanding of how these drugs affect signaling pathways as well as associated diseases.

Calmodulin (CaM) is a calcium-sensing protein which makes other proteins dependent on the surrounding calcium concentration by binding to these proteins. Such protein-protein interactions with CaM are vital for calcium to control many physiological pathways within the cell. The antipsychotic drug trifluoperazine (TFP) inhibits CaM’s ability to bind and regulate other proteins. Here, we use molecular dynamics simulations together with Markov state modeling and machine learning to understand the structural and dynamical features by which TFP bound to the one domain of CaM prevents association to other proteins. We find that TFP encourages CaM to adopt a conformation that is like the one stabilized in absence of calcium: depending on the binding orientation of TFP, the drug indeed either affects the CaM hydrophobic binding pocket, the calcium binding sites or the secondary structure content in the domain. Understanding TFP binding is a first step towards designing better drugs targeting CaM.

Allostery: Allosteric Cancer Drivers and Innovative Allosteric Drugs

Here, we discuss the principles of allosteric activating mutations, propagation downstream of the signals that they prompt, and allosteric drugs, with examples from the Ras signaling network. We focus on Abl kinase where mutations shift the landscape toward the active, imatinib binding-incompetent conformation, likely resulting in the high affinity ATP outcompeting drug binding. Recent pharmacological innovation extends to allosteric inhibitor (GNF-5)-linked PROTAC, targeting Bcr-Abl1 myristoylation site, and broadly, allosteric heterobifunctional degraders that destroy targets, rather than inhibiting them. Designed chemical linkers in bifunctional degraders can connect the allosteric ligand that binds the target protein and the E3 ubiquitin ligase warhead anchor. The physical properties and favored conformational state of the engineered linker can precisely coordinate the distance and orientation between the target and the recruited E3. Allosteric PROTACs, noncompetitive molecular glues, and bitopic ligands, with covalent links of allosteric ligands and orthosteric warheads, increase the effective local concentration of productively oriented and placed ligands. Through covalent chemical or peptide linkers, allosteric drugs can collaborate with competitive drugs, degrader anchors, or other molecules of choice, driving innovative drug discovery.

Characterizing Metastable States with the Help of Machine Learning

Present-day atomistic simulations generate long trajectories of ever more complex systems. Analyzing these data, discovering metastable states, and uncovering their nature are becoming increasingly challenging. In this paper, we first use the variational approach to conformation dynamics to discover the slowest dynamical modes of the simulations. This allows the different metastable states of the system to be located and organized hierarchically. The physical descriptors that characterize metastable states are discovered by means of a machine learning method. We show in the cases of two proteins, chignolin and bovine pancreatic trypsin inhibitor, how such analysis can be effortlessly performed in a matter of seconds. Another strength of our approach is that it can be applied to the analysis of both unbiased and biased simulations.

How Fragile We Are: Influence of Stimulator of Interferon Genes (STING) Variants on Pathogen Recognition and Immune Response Efficiency

The stimulator of interferon genes (STING) protein is a cornerstone of the human immune response. Its activation by cGAMP in the presence of cytosolic DNA stimulates the production of type I interferons and inflammatory cytokines. In the human population, several STING variants exist and exhibit dramatic differences in their activity, impacting the efficiency of the host defense against infections. Understanding the molecular mechanisms of these variants opens perspectives for personalized medicine treatments against diseases such as viral infections, cancers, or autoinflammatory diseases. Through microsecond-scale molecular modeling simulations, contact analyses, and machine learning techniques, we reveal the dynamic behavior of four STING variants (wild type, G230A, R293Q, and G230A/R293Q) and rationalize the variability of efficiency observed experimentally. Our results show that the decrease in STING activity is linked to a stiffening of key structural elements of the binding cavity together with changes in the interaction patterns within the protein.

Novel insights into the structural changes induced by disease-associated mutations in TDP-43: a computational approach

Over the last two decades, the pathogenic aggregation of TAR DNA-binding protein 43 (TDP-43) is found to be strongly associated with several fatal neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTD), etc. While the mutations and truncation in TDP-43 protein have been suggested to be responsible for TDP-43 pathogenesis by accelerating the aggregation process, the effects of these mutations on the bio-mechanism of pathological TDP-43 protein remained poorly understood. Investigating this at the molecular level, we formulized an integrated workflow of molecular dynamic simulation and machine learning models (MD-ML). By performing an extensive structural analysis of three disease-related mutations (i.e., I168A, D169G, and I168A-D169G) in the conserved RNA recognition motifs (RRM1) of TDP-43, we observed that the I168A-D169G double mutant delineates the highest packing of the protein inner core as compared to the other mutations, which may indicate more stability and higher chances of pathogenesis. Moreover, through our MD-ML workflow, we identified the biological descriptors of TDP-43 which includes the interacting residue pairs and individual protein residues that influence the stability of the protein and could be experimentally evaluated to develop potential therapeutic strategies.Communicated by Ramaswamy H. Sarma.

Specific Recognition of the 5'-Untranslated Region of West Nile Virus Genome by Human Innate Immune System

In the last few years, the sudden outbreak of COVID-19 caused by SARS-CoV-2 proved the crucial importance of understanding how emerging viruses work and proliferate, in order to avoid the repetition of such a dramatic sanitary situation with unprecedented social and economic costs. West Nile Virus is a mosquito-borne pathogen that can spread to humans and induce severe neurological problems. This RNA virus caused recent remarkable outbreaks, notably in Europe, highlighting the need to investigate the molecular mechanisms of its infection process in order to design and propose efficient antivirals. Here, we resort to all-atom Molecular Dynamics simulations to characterize the structure of the 5′-untranslated region of the West Nile Virus genome and its specific recognition by the human innate immune system via oligoadenylate synthetase. Our simulations allowed us to map the interaction network between the viral RNA and the host protein, which drives its specific recognition and triggers the host immune response. These results may provide fundamental knowledge that can assist further antivirals’ design, including therapeutic RNA strategies.

Autophagy and evasion of the immune system by SARS-CoV-2. Structural features of the non-structural protein 6 from wild type and Omicron viral strains interacting with a model lipid bilayer

The viral cycle of SARS-CoV-2 is based on a complex interplay with the cellular machinery, which is mediated by specific proteins eluding or hijacking the cellular defense mechanisms. Among the complex pathways induced by the viral infection, autophagy is particularly crucial and is strongly influenced by the action of the non-structural protein 6 (Nsp6) interacting with the endoplasmic reticulum membrane. Importantly, differently from other non-structural proteins, Nsp6 is mutated in the recently emerged Omicron variant, suggesting a possible different role of autophagy. In this contribution we explore, for the first time, the structural properties of Nsp6 thanks to long-timescale molecular dynamics simulations and machine learning analysis, identifying the interaction patterns with the lipid membrane. We also show how the mutation brought by the Omicron variant may indeed modify some of the specific interactions, and more particularly help anchor the viral protein to the lipid bilayer interface.

The viral cycle of SARS-CoV-2 is based on a complex interplay with the cellular machinery, which is mediated by specific proteins eluding or hijacking the cellular defense mechanisms.

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.

An Interpretable Convolutional Neural Network Framework for Analyzing Molecular Dynamics Trajectories: a Case Study on Functional States for G-Protein-Coupled Receptors

Molecular dynamics (MD) simulations have made great contribution to revealing structural and functional mechanisms for many biomolecular systems. However, how to identify functional states and important residues from vast conformation space generated by MD remains challenging; thus an intelligent navigation is highly desired. Despite intelligent advantages of deep learning exhibited in analyzing MD trajectory, its black-box nature limits its application. To address this problem, we explore an interpretable convolutional neural network (CNN)-based deep learning framework to automatically identify diverse active states from the MD trajectory for G-protein-coupled receptors (GPCRs), named the ICNNMD model. To avoid the information loss in representing the conformation structure, the pixel representation is introduced, and then the CNN module is constructed to efficiently extract features followed by a fully connected neural network to realize the classification task. More importantly, we design a local interpretable model-agnostic explanation interpreter for the classification result by local approximation with a linear model, through which important residues underlying distinct active states can be quickly identified. Our model showcases higher than 99% classification accuracy for three important GPCR systems with diverse active states. Notably, some important residues in regulating different biased activities are successfully identified, which are beneficial to elucidating diverse activation mechanisms for GPCRs. Our model can also serve as a general tool to analyze MD trajectory for other biomolecular systems. All source codes are freely available at https://github.com/Jane-Liu97/ICNNMD for aiding MD studies.

GLOW: A Workflow Integrating Gaussian-Accelerated Molecular Dynamics and Deep Learning for Free Energy Profiling

We introduce a Gaussian-accelerated molecular dynamics (GaMD), deep learning (DL), and free energy profiling workflow (GLOW) to predict molecular determinants and map free energy landscapes of biomolecules. All-atom GaMD-enhanced sampling simulations are first performed on biomolecules of interest. Structural contact maps are then calculated from GaMD simulation frames and transformed into images for building DL models using a convolutional neural network. Important structural contacts are further determined from DL models of attention maps of the structural contact gradients, which allow us to identify the system reaction coordinates. Finally, free energy profiles are calculated for the selected reaction coordinates through energetic reweighting of the GaMD simulations. We have also successfully demonstrated GLOW for the characterization of activation and allosteric modulation of a G protein-coupled receptor, using the adenosine A₁ receptor (A₁AR) as a model system. GLOW findings are highly consistent with previous experimental and computational studies of the A₁AR, while also providing further mechanistic insights into the receptor function. In summary, GLOW provides a systematic approach to mapping free energy landscapes of biomolecules. The GLOW workflow and its user manual can be downloaded at http://miaolab.org/GLOW.

Identification of electroporation sites in the complex lipid organization of the plasma membrane

The plasma membrane of a biological cell is a complex assembly of lipids and membrane proteins, which tightly regulate transmembrane transport. When a cell is exposed to strong electric field, the membrane integrity becomes transiently disrupted by formation of transmembrane pores. This phenomenon termed electroporation is already utilized in many rapidly developing applications in medicine including gene therapy, cancer treatment, and treatment of cardiac arrhythmias. However, the molecular mechanisms of electroporation are not yet sufficiently well understood; in particular, it is unclear where exactly pores form in the complex organization of the plasma membrane. In this study, we combine coarse-grained molecular dynamics simulations, machine learning methods, and Bayesian survival analysis to identify how formation of pores depends on the local lipid organization. We show that pores do not form homogeneously across the membrane, but colocalize with domains that have specific features, the most important being high density of polyunsaturated lipids. We further show that knowing the lipid organization is sufficient to reliably predict poration sites with machine learning. Additionally, by analysing poration kinetics with Bayesian survival analysis we show that poration does not depend solely on local lipid arrangement, but also on membrane mechanical properties and the polarity of the electric field. Finally, we discuss how the combination of atomistic and coarse-grained molecular dynamics simulations, machine learning methods, and Bayesian survival analysis can guide the design of future experiments and help us to develop an accurate description of plasma membrane electroporation on the whole-cell level. Achieving this will allow us to shift the optimization of electroporation applications from blind trial-and-error approaches to mechanistic-driven design.

Machine Learning at the Interface of Polymer Science and Biology: How Far Can We Go?

This Perspective outlines recent progress and future directions for using machine learning (ML), a data-driven method, to address critical questions in the design, synthesis, processing, and characterization of biomacromolecules. The achievement of these tasks requires the navigation of vast and complex chemical and biological spaces, difficult to accomplish with reasonable speed. Using modern algorithms and supercomputers, quantum physics methods are able to examine systems containing a few hundred interacting species and determine the probability of finding them in a particular region of phase space, thereby anticipating their properties. Likewise, modern approaches in chemistry and biomolecular simulation, supported by high performance computing, have culminated in producing data sets of escalating size and intrinsically high complexity. Hence, using ML to extract relevant information from these fields is of paramount importance to advance our understanding of chemical and biomolecular systems. At the heart of ML approaches lie statistical algorithms, which by evaluating a portion of a given data set, identify, learn, and manipulate the underlying rules that govern the whole data set. The assembly of a quality model to represent the data followed by the predictions and elimination of error sources are the key steps in ML. In addition to a growing infrastructure of ML tools to address complex problems, an increasing number of aspects related to our understanding of the fundamental properties of biomacromolecules are exposed to ML. These fields, including those residing at the interface of polymer science and biology (i.e., structure determination, de novo design, folding, and dynamics), strive to adopt and take advantage of the transformative power offered by approaches in the ML domain, which clearly has the potential of accelerating research in the field of biomacromolecules.

Is There a Need for a More Precise Description of Biomolecule Interactions to Understand Cell Function?

An important goal of biological research is to explain and hopefully predict cell behavior from the molecular properties of cellular components. Accordingly, much work was done to build extensive “omic” datasets and develop theoretical methods, including computer simulation and network analysis to process as quantitatively as possible the parameters contained in these resources. Furthermore, substantial effort was made to standardize data presentation and make experimental results accessible to data scientists. However, the power and complexity of current experimental and theoretical tools make it more and more difficult to assess the capacity of gathered parameters to support optimal progress in our understanding of cell function. The purpose of this review is to focus on biomolecule interactions, the interactome, as a specific and important example, and examine the limitations of the explanatory and predictive power of parameters that are considered as suitable descriptors of molecular interactions. Recent experimental studies on important cell functions, such as adhesion and processing of environmental cues for decision-making, support the suggestion that it should be rewarding to complement standard binding properties such as affinity and kinetic constants, or even force dependence, with less frequently used parameters such as conformational flexibility or size of binding molecules.

Allosteric Effect of Nanobody Binding on Ligand-Specific Active States of the β2 Adrenergic Receptor

Nanobody binding stabilizes G-protein-coupled receptors (GPCR) in a fully active state and modulates their affinity for bound ligands. However, the atomic-level basis for this allosteric regulation remains elusive. Here, we investigate the conformational changes induced by the binding of a nanobody (Nb80) on the active-like β2 adrenergic receptor (β2AR) via enhanced sampling molecular dynamics simulations. Dimensionality reduction analysis shows that Nb80 stabilizes structural features of the β2AR with an ∼14 Å outward movement of transmembrane helix 6 and a close proximity of transmembrane (TM) helices 5 and 7, and favors the fully active-like conformation of the receptor, independent of ligand binding, in contrast to the conditions under which no intracellular binding partner is bound, in which case the receptor is only stabilized in an intermediate-active state. This activation is supported by the residues located at hotspots located on TMs 5, 6, and 7, as shown by supervised machine learning methods. Besides, ligand-specific subtle differences in the conformations assumed by intracellular loop 2 and extracellular loop 2 are captured from the trajectories of various ligand-bound receptors in the presence of Nb80. Dynamic network analysis further reveals that Nb80 binding triggers tighter and stronger local communication networks between the Nb80 and the ligand-binding sites, primarily involving residues around ICL2 and the intracellular end of TM3, TM5, TM6, as well as ECL2, ECL3, and the extracellular ends of TM6 and TM7. In particular, we identify unique allosteric signal transmission mechanisms between the Nb80-binding site and the extracellular domains in conformations modulated by a full agonist, BI167107, and a G-protein-biased partial agonist, salmeterol, involving mainly TM1 and TM2, and TM5, respectively. Altogether, our results provide insights into the effect of intracellular binding partners on the GPCR activation mechanism, which should be taken into account in structure-based drug discovery.

Resolving Dynamics in the Ensemble: Finding Paths through Intermediate States and Disordered Protein Structures

Proteins have been found to inhabit a diverse set of three-dimensional structures. The dynamics that govern protein interconversion between structures happen over a wide range of time scales─picoseconds to seconds. Our understanding of protein functions and dynamics is largely reliant upon our ability to elucidate physically populated structures. From an experimental structural characterization perspective, we are often limited to measuring the ensemble-averaged structure both in the steady-state and time-resolved regimes. Generating kinetic models and understanding protein structure-function relationships require atomistic knowledge of the populated states in the ensemble. In this Perspective, we present ensemble refinement methodologies that integrate time-resolved experimental signals with molecular dynamics models. We first discuss integration of experimental structural restraints to molecular models in disordered protein systems that adhere to the principle of maximum entropy for creating a complete set of ensemble structures. We then propose strategies to find kinetic pathways between the refined structures, using time-resolved inputs to guide molecular dynamics trajectories and the use of inference to generate tailored stimuli to prepare a desired ensemble of protein states.

Mechanotransduction as a major driver of cell behaviour: mechanisms, and relevance to cell organization and future research

How do cells process environmental cues to make decisions? This simple question is still generating much experimental and theoretical work, at the border of physics, chemistry and biology, with strong implications in medicine. The purpose of mechanobiology is to understand how biochemical and physical cues are turned into signals through mechanotransduction. Here, we review recent evidence showing that (i) mechanotransduction plays a major role in triggering signalling cascades following cell-neighbourhood interaction; (ii) the cell capacity to continually generate forces, and biomolecule properties to undergo conformational changes in response to piconewton forces, provide a molecular basis for understanding mechanotransduction; and (iii) mechanotransduction shapes the guidance cues retrieved by living cells and the information flow they generate. This includes the temporal and spatial properties of intracellular signalling cascades. In conclusion, it is suggested that the described concepts may provide guidelines to define experimentally accessible parameters to describe cell structure and dynamics, as a prerequisite to take advantage of recent progress in high-throughput data gathering, computer simulation and artificial intelligence, in order to build a workable, hopefully predictive, account of cell signalling networks.

Molecular Mechanisms Associated with Clustered Lesion-Induced Impairment of 8-oxoG Recognition by the Human Glycosylase OGG1

The 8-oxo-7,8-dihydroguanine, referred to as 8-oxoG, is a highly mutagenic DNA lesion that can provoke the appearance of mismatches if it escapes the DNA Damage Response. The specific recognition of its structural signature by the hOGG1 glycosylase is the first step along the Base Excision Repair pathway, which ensures the integrity of the genome by preventing the emergence of mutations. 8-oxoG formation, structural features, and repair have been matters of extensive research; more recently, this active field of research expended to the more complicated case of 8-oxoG within clustered lesions. Indeed, the presence of a second lesion within 1 or 2 helix turns can dramatically impact the repair yields of 8-oxoG by glycosylases. In this work, we use μs-range molecular dynamics simulations and machine-learning-based postanalysis to explore the molecular mechanisms associated with the recognition of 8-oxoG by hOGG1 when embedded in a multiple-lesion site with a mismatch in 5′ or 3′. We delineate the stiffening of the DNA–protein interactions upon the presence of the mismatches, and rationalize the much lower repair yields reported with a 5′ mismatch by describing the perturbation of 8-oxoG structural features upon addition of an adjacent lesion.

AI-driven prediction of SARS-CoV-2 variant binding trends from atomistic simulations

We present a novel technique to predict binding affinity trends between two molecules from atomistic molecular dynamics simulations. The technique uses a neural network algorithm applied to a series of images encoding the distance between two molecules in time. We demonstrate that our algorithm is capable of separating with high accuracy non-hydrophobic mutations with low binding affinity from those with high binding affinity. Moreover, we show high accuracy in prediction using a small subset of the simulation, therefore requiring a much shorter simulation time. We apply our algorithm to the binding between several variants of the SARS-CoV-2 spike protein and the human receptor ACE2.

Risk Prediction of Cardiovascular Events by Exploration of Molecular Data with Explainable Artificial Intelligence

Cardiovascular diseases (CVD) annually take almost 18 million lives worldwide. Most lethal events occur months or years after the initial presentation. Indeed, many patients experience repeated complications or require multiple interventions (recurrent events). Apart from affecting the individual, this leads to high medical costs for society. Personalized treatment strategies aiming at prediction and prevention of recurrent events rely on early diagnosis and precise prognosis. Complementing the traditional environmental and clinical risk factors, multi-omics data provide a holistic view of the patient and disease progression, enabling studies to probe novel angles in risk stratification. Specifically, predictive molecular markers allow insights into regulatory networks, pathways, and mechanisms underlying disease. Moreover, artificial intelligence (AI) represents a powerful, yet adaptive, framework able to recognize complex patterns in large-scale clinical and molecular data with the potential to improve risk prediction. Here, we review the most recent advances in risk prediction of recurrent cardiovascular events, and discuss the value of molecular data and biomarkers for understanding patient risk in a systems biology context. Finally, we introduce explainable AI which may improve clinical decision systems by making predictions transparent to the medical practitioner.

Machine Learning Reveals the Critical Interactions for SARS-CoV-2 Spike Protein Binding to ACE2

SARS-CoV and SARS-CoV-2 bind to the human ACE2 receptor in practically identical conformations, although several residues of the receptor-binding domain (RBD) differ between them. Herein, we have used molecular dynamics (MD) simulations, machine learning (ML), and free-energy perturbation (FEP) calculations to elucidate the differences in binding by the two viruses. Although only subtle differences were observed from the initial MD simulations of the two RBD-ACE2 complexes, ML identified the individual residues with the most distinctive ACE2 interactions, many of which have been highlighted in previous experimental studies. FEP calculations quantified the corresponding differences in binding free energies to ACE2, and examination of MD trajectories provided structural explanations for these differences. Lastly, the energetics of emerging SARS-CoV-2 mutations were studied, showing that the affinity of the RBD for ACE2 is increased by N501Y and E484K mutations but is slightly decreased by K417N.

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

Understanding the structural determinants of a protein's biochemical properties, such as activity and stability, is a major challenge in biology and medicine. Comparing computer simulations of protein variants with different biochemical properties is an increasingly powerful means to drive progress. However, success often hinges on dimensionality reduction algorithms for simplifying the complex ensemble of structures each variant adopts. Unfortunately, common algorithms rely on potentially misleading assumptions about what structural features are important, such as emphasizing larger geometric changes over smaller ones. Here we present DiffNets, self-supervised autoencoders that avoid such assumptions, and automatically identify the relevant features, by requiring that the low-dimensional representations they learn are sufficient to predict the biochemical differences between protein variants. For example, DiffNets automatically identify subtle structural signatures that predict the relative stabilities of β-lactamase variants and duty ratios of myosin isoforms. DiffNets should also be applicable to understanding other perturbations, such as ligand binding.

Recognition of a tandem lesion by DNA bacterial formamidopyrimidine glycosylases explored combining molecular dynamics and machine learning

The combination of several closely spaced DNA lesions, which can be induced by a single radical hit, constitutes a hallmark in the DNA damage landscape and radiation chemistry. The occurrence of such a tandem base lesion gives rise to a strong coupling with the double helix degrees of freedom and induces important structural deformations, in contrast to DNA strands containing a single oxidized nucleobase. Although such complex lesions are known to be refractory to repair by DNA glycosylases, there is still a lack of structural evidence to rationalize these phenomena. In this contribution, we explore, by numerical modeling and molecular simulations, the behavior of the bacterial glycosylase responsible for base excision repair (MutM), specialized in excising oxidatively-damaged defects such as 7,8-dihydro-8-oxoguanine (8-oxoG). The difference in lesion recognition between a simple damage and a tandem lesion featuring an additional abasic site is assessed at atomistic resolution owing to microsecond molecular dynamics simulations and machine learning postprocessing, allowing to extensively pinpoint crucial differences in the interaction patterns of the damaged bases. Our results reveal substantial changes in the interaction network surrounding the 8-oxoG upon addition of an adjacent abasic site, leading to the perturbation of the intercalation triad which is crucial for lesion recognition and processing. The recognition process might also be impacted by a more constrained MutM-DNA binding upon tandem damage, as shown by the machine learning post-processing. This work advocates for the use of such high throughput numerical simulations for exploring the complex combinatorial chemistry of tandem DNA lesions repair and more generally local multiple damaged sites of the utmost significance in radiation chemistry.

Computational studies of the mitochondrial carrier family SLC25. Present status and future perspectives

The members of the mitochondrial carrier family, also known as solute carrier family 25 (SLC25), are transmembrane proteins involved in the translocation of a plethora of small molecules between the mitochondrial intermembrane space and the matrix. These transporters are characterized by three homologous domains structure and a transport mechanism that involves the transition between different conformations. Mutations in regions critical for these transporters’ function often cause several diseases, given the crucial role of these proteins in the mitochondrial homeostasis. Experimental studies can be problematic in the case of membrane proteins, in particular concerning the characterization of the structure–function relationships. For this reason, computational methods are often applied in order to develop new hypotheses or to support/explain experimental evidence. Here the computational analyses carried out on the SLC25 members are reviewed, describing the main techniques used and the outcome in terms of improved knowledge of the transport mechanism. Potential future applications on this protein family of more recent and advanced in silico methods are also suggested.

Predicting new protein conformations from molecular dynamics simulation conformational landscapes and machine learning

Molecular dynamics (MD) simulations are a popular method of studying protein structure and function, but are unable to reliably sample all relevant conformational space in reasonable computational timescales. A range of enhanced sampling methods are available that can improve conformational sampling, but these do not offer a complete solution. We present here a proof-of-principle method of combining MD simulation with machine learning to explore protein conformational space. An autoencoder is used to map snapshots from MD simulations onto a user-defined conformational landscape defined by principal components analysis or specific structural features, and we show that we can predict, with useful accuracy, conformations that are not present in the training data. This method offers a new approach to the prediction of new low energy/physically realistic structures of conformationally dynamic proteins and allows an alternative approach to enhanced sampling of MD simulations.

Identification of ligand-specific G protein-coupled receptor states and prediction of downstream efficacy via data-driven modeling

Ligand binding stabilizes different G protein-coupled receptor states via a complex allosteric process that is not completely understood. Here, we have derived free energy landscapes describing activation of the β2 adrenergic receptor bound to ligands with different efficacy profiles using enhanced sampling molecular dynamics simulations. These reveal shifts toward active-like states at the Gprotein-binding site for receptors bound to partial and full agonists, and that the ligands modulate the conformational ensemble of the receptor by tuning protein microswitches. We indeed find an excellent correlation between the conformation of the microswitches close to the ligand binding site and in the transmembrane region and experimentally reported cyclic adenosine monophosphate signaling responses. Dimensionality reduction further reveals the similarity between the unique conformational states induced by different ligands, and examining the output of classifiers highlights two distant hotspots governing agonism on transmembrane helices 5 and 7.

Molecular Dynamics Approach for Capturing Calixarene-Protein Interactions: The Case of Cytochrome C

Functionalized supramolecular cages are of growing importance in biology and biochemistry. They have recently been proposed as efficient auxiliaries to obtain high-resolution cocrystallized proteins. Here, we propose a molecular dynamics investigation of the supramolecular association of sulfonated calix-[8]-arenes to cytochrome c starting from initially distant proteins and ligands. We characterize two main binding sites for the sulfonated calixarene on the cytochrome c surface which are in perfect agreement with the previous experiments with regard to the structure (comparison with the X-ray structure PDB 6GD8) and the binding free energies [comparison between the molecular mechanics Poisson-Boltzmann surface area analysis and the isothermal titration calorimetry measurements]. The per-residue decomposition of the interaction energies reveals the detailed picture of this electrostatically driven association and notably the role of arginine R13 as a bridging residue between the two main anchoring sites. In addition, the analysis of the residue behavior by means of a supervised machine learning protocol unveils the formation of a hydrogen bond network far from the binding sites, increasing the rigidity of the protein. This study paves the way toward an automated procedure to predict the supramolecular protein-cage association, with the possibility of a computational screening of new promising derivatives for controlled protein assembly and protein surface recognition processes.

Nucleosomal embedding reshapes the dynamics of abasic sites

Apurinic/apyrimidinic (AP) sites are the most common DNA lesions, which benefit from a most efficient repair by the base excision pathway. The impact of losing a nucleobase on the conformation and dynamics of B-DNA is well characterized. Yet AP sites seem to present an entirely different chemistry in nucleosomal DNA, with lifetimes reduced up to 100-fold, and the much increased formation of covalent DNA-protein cross-links leading to strand breaks, refractory to repair. We report microsecond range, all-atom molecular dynamics simulations that capture the conformational dynamics of AP sites and their tetrahydrofuran analogs at two symmetrical positions within a nucleosome core particle, starting from a recent crystal structure. Different behaviours between the deoxyribo-based and tetrahydrofuran-type abasic sites are evidenced. The two solvent-exposed lesion sites present contrasted extrahelicities, revealing the crucial role of the position of a defect around the histone core. Our all-atom simulations also identify and quantify the frequency of several spontaneous, non-covalent interactions between AP and positively-charged residues from the histones H2A and H2B tails that prefigure DNA-protein cross-links. Such an in silico mapping of DNA-protein cross-links gives important insights for further experimental studies involving mutagenesis and truncation of histone tails to unravel mechanisms of DPCs formation.

Impact of the Nucleosome Histone Core on the Structure and Dynamics of DNA-Containing Pyrimidine-Pyrimidone (6-4) Photoproduct

The pyrimidine-pyrimidone (6-4) photoproduct (64-PP) is an important photoinduced DNA lesion constituting a mutational signature for melanoma. The structural impact of 64-PP on DNA complexed with histones affects the lesion mutagenicity and repair but remains poorly understood. Here we investigate the conformational dynamics of DNA-containing 64-PP within the nucleosome core particle by atomic-resolution molecular dynamics simulations and multiscale data analysis. We demonstrate that the histone core exerts important mechanical restraints that largely decrease global DNA structural fluctuations. However, the local DNA flexibility at the damaged site is enhanced due to imperfect structural adaptation to restraints imposed by the histone core. If 64-PP faces the histone core and is therefore not directly accessible by the repair protein, the complementary strand facing the solvent is deformed and exhibits higher flexibility than the corresponding strand in a naked, undamaged DNA. This may serve as an initial recognition signal for repair. Our simulations also pinpoint the structural role of proximal residues from the truncated histone tails.

Integrating NMR and simulations reveals motions in the UUCG tetraloop

We provide an atomic-level description of the structure and dynamics of the UUCG RNA stem-loop by combining molecular dynamics simulations with experimental data. The integration of simulations with exact nuclear Overhauser enhancements data allowed us to characterize two distinct states of this molecule. The most stable conformation corresponds to the consensus three-dimensional structure. The second state is characterized by the absence of the peculiar non-Watson-Crick interactions in the loop region. By using machine learning techniques we identify a set of experimental measurements that are most sensitive to the presence of non-native states. We find that although our MD ensemble, as well as the consensus UUCG tetraloop structures, are in good agreement with experiments, there are remaining discrepancies. Together, our results show that (i) the MD simulation overstabilize a non-native loop conformation, (ii) eNOE data support its presence with a population of ≈10% and (iii) the structural interpretation of experimental data for dynamic RNAs is highly complex, even for a simple model system such as the UUCG tetraloop.

A Two-Ended Data-Driven Accelerated Sampling Method for Exploring the Transition Pathways between Two Known States of Protein

Conformational transitions of protein between different states are often associated with their biological functions. These dynamic processes, however, are usually not easy to be well characterized by experimental measurements, mainly because of inadequate temporal and spatial resolution. Meantime, sampling of configuration space with molecular dynamics (MD) simulations is still a challenge. Here we proposed a robust two-ended data-driven accelerated (teDA2) conformational sampling method, which drives the structural change in an adaptively updated feature space without introducing a bias potential. teDA2 was applied to explore adenylate kinase (ADK), a model with well characterized "open" and "closed" states. A single conformational transition event of ADK could be achieved within only a few or tens of nanoseconds sampled with teDA2. By analyzing hundreds of transition events, we reproduced different mechanisms and the associated pathways for domain motion of ADK reported in the literature. The multiroute characteristic of ADK was confirmed by the fact that some metastable states identified with teDA2 resemble available crystal structures determined at different conditions. This feature was further validated with Markov state modeling with independent MD simulations. Therefore, our work provides strong evidence for the conformational plasticity of protein, which is mainly due to the inherent degree of flexibility. As a reliable and efficient enhanced sampling protocol, teDA2 could be used to study the dynamics between functional states of various biomolecular machines.

Energy Landscapes Reveal Agonist Control of G Protein-Coupled Receptor Activation via Microswitches

Agonist binding to G protein-coupled receptors (GPCRs) leads to conformational changes in the transmembrane region that activate cytosolic signaling pathways. Although high-resolution structures of different receptor states are available, atomistic details of allosteric signaling across the membrane remain elusive. We calculated free energy landscapes of β₂ adrenergic receptor activation using atomistic molecular dynamics simulations in an optimized string of swarms framework, which shed new light on how microswitches govern the equilibrium between conformational states. Contraction of the extracellular binding site in the presence of the agonist BI-167107 is obligatorily coupled to conformational changes in a connector motif located in the core of the transmembrane region. The connector is probabilistically coupled to the conformation of the intracellular region. An active connector promotes desolvation of a buried cavity, a twist of the conserved NPxxY motif, and an interaction between two conserved tyrosines in transmembrane helices 5 and 7 (Y–Y motif), which lead to a larger population of active-like states at the G protein binding site. This coupling is augmented by protonation of the strongly conserved Asp79^2.50. The agonist binding site hence communicates with the intracellular region via a cascade of locally connected microswitches. Characterization of these can be used to understand how ligands stabilize distinct receptor states and contribute to development drugs with specific signaling properties. The developed simulation protocol can likely be transferred to other class A GPCRs.