Unsupervised and supervised AI on molecular dynamics simulations reveals complex characteristics of HLA-A2-peptide immunogenicity.

Immunologic recognition of peptide antigens bound to class I major histocompatibility complex (MHC) molecules is essential to both novel immunotherapeutic development and human health at large. Current methods for predicting antigen peptide immunogenicity rely primarily on simple sequence representations, which allow for some understanding of immunogenic features but provide inadequate consideration of the full scale of molecular mechanisms tied to peptide recognition. We here characterize contributions that unsupervised and supervised artificial intelligence (AI) methods can make toward understanding and predicting MHC(HLA-A2)-peptide complex immunogenicity when applied to large ensembles of molecular dynamics simulations. We first show that an unsupervised AI method allows us to identify subtle features that drive immunogenicity differences between a cancer neoantigen and its wild-type peptide counterpart. Next, we demonstrate that a supervised AI method for class I MHC(HLA-A2)-peptide complex classification significantly outperforms a sequence model on small datasets corrected for trivial sequence correlations. Furthermore, we show that both unsupervised and supervised approaches reveal determinants of immunogenicity based on time-dependent molecular fluctuations and anchor position dynamics outside the MHC binding groove. We discuss implications of these structural and dynamic immunogenicity correlates for the induction of T cell responses and therapeutic T cell receptor design.

In silico design and validation of high-affinity RNA aptamers targeting epithelial cellular adhesion molecule dimers.

Nucleic acid aptamers hold great promise for therapeutic applications due to their favorable intrinsic properties, as well as high-throughput experimental selection techniques. Despite the utility of the systematic evolution of ligands by the exponential enrichment (SELEX) method for aptamer determination, complementary in silico aptamer design is highly sought after to facilitate virtual screening and increased understanding of important nucleic acid-protein interactions. Here, with a combined experimental and theoretical approach, we have developed two optimal epithelial cellular adhesion molecule (EpCAM) aptamers. Our structure-based in silico method first predicts their binding modes and then optimizes them for EpCAM with molecular dynamics simulations, docking, and free energy calculations. Our isothermal titration calorimetry experiments further confirm that the EpCAM aptamers indeed exhibit enhanced affinity over a previously patented nanomolar aptamer, EP23. Moreover, our study suggests that EP23 and the de novo designed aptamers primarily bind to EpCAM dimers (and not monomers, as hypothesized in previous published works), suggesting a paradigm for developing EpCAM-targeted therapies.

Analysis of Training and Seed Bias in Small Molecules Generated with a Conditional Graph-Based Variational Autoencoder─Insights for Practical AI-Driven Molecule Generation.

The application of deep learning to generative molecule design has shown early promise for accelerating lead series development. However, questions remain concerning how factors like training, data set, and seed bias impact the technology's utility to medicinal and computational chemists. In this work, we analyze the impact of seed and training bias on the output of an activity-conditioned graph-based variational autoencoder (VAE). Leveraging a massive, labeled data set corresponding to the dopamine D2 receptor, our graph-based generative model is shown to excel in producing desired conditioned activities and favorable unconditioned physical properties in generated molecules. We implement an activity-swapping method that allows for the activation, deactivation, or retention of activity of molecular seeds, and we apply independent deep learning classifiers to verify the generative results. Overall, we uncover relationships between noise, molecular seeds, and training set selection across a range of latent-space sampling procedures, providing important insights for practical AI-driven molecule generation.

Simplified, interpretable graph convolutional neural networks for small molecule activity prediction.

We here present a streamlined, explainable graph convolutional neural network (gCNN) architecture for small molecule activity prediction. We first conduct a hyperparameter optimization across nearly 800 protein targets that produces a simplified gCNN QSAR architecture, and we observe that such a model can yield performance improvements over both standard gCNN and RF methods on difficult-to-classify test sets. Additionally, we discuss how reductions in convolutional layer dimensions potentially speak to the "anatomical" needs of gCNNs with respect to radial coarse graining of molecular substructure. We augment this simplified architecture with saliency map technology that highlights molecular substructures relevant to activity, and we perform saliency analysis on nearly 100 data-rich protein targets. We show that resultant substructural clusters are useful visualization tools for understanding substructure-activity relationships. We go on to highlight connections between our models' saliency predictions and observations made in the medicinal chemistry literature, focusing on four case studies of past lead finding and lead optimization campaigns.

T cell receptors for the HIV KK10 epitope from patients with differential immunologic control are functionally indistinguishable.

HIV controllers (HCs) are individuals who can naturally control HIV infection, partially due to potent HIV-specific CD8+ T cell responses. Here, we examined the hypothesis that superior function of CD8+ T cells from HCs is encoded by their T cell receptors (TCRs). We compared the functional properties of immunodominant HIV-specific TCRs obtained from HLA-B*2705 HCs and chronic progressors (CPs) following expression in primary T cells. T cells transduced with TCRs from HCs and CPs showed equivalent induction of epitope-specific cytotoxicity, cytokine secretion, and antigen-binding properties. Transduced T cells comparably, albeit modestly, also suppressed HIV infection in vitro and in humanized mice. We also performed extensive molecular dynamics simulations that provided a structural basis for similarities in cytotoxicity and epitope cross-reactivity. These results demonstrate that the differential abilities of HIV-specific CD8+ T cells from HCs and CPs are not genetically encoded in the TCRs alone and must depend on additional factors.

Combining Docking Pose Rank and Structure with Deep Learning Improves Protein-Ligand Binding Mode Prediction over a Baseline Docking Approach.

We present a simple, modular graph-based convolutional neural network that takes structural information from protein-ligand complexes as input to generate models for activity and binding mode prediction. Complex structures are generated by a standard docking procedure and fed into a dual-graph architecture that includes separate subnetworks for the ligand bonded topology and the ligand-protein contact map. Recent work has indicated that data set bias drives many past promising results derived from combining deep learning and docking. Our dual-graph network allows contributions from ligand identity that give rise to such biases to be distinguished from effects of protein-ligand interactions on classification. We show that our neural network is capable of learning from protein structural information when, as in the case of binding mode prediction, an unbiased data set is constructed. We next develop a deep learning model for binding mode prediction that uses docking ranking as input in combination with docking structures. This strategy mirrors past consensus models and outperforms a baseline docking program (AutoDock Vina) in a variety of tests, including on cross-docking data sets that mimic real-world docking use cases. Furthermore, the magnitudes of network predictions serve as reliable measures of model confidence.

Rare Dissipative Transitions Punctuate the Initiation of Chemical Denaturation in Proteins.

Protein unfolding dynamics are bound by their degree of entropy production, a quantity that relates the amount of heat dissipated by a nonequilibrium process to a system's forward and time-reversed trajectories. We here explore the statistics of heat dissipation that emerge in protein molecules subjected to a chemical denaturant. Coupling large molecular dynamics datasets and Markov state models with the theory of entropy production, we demonstrate that dissipative processes can be rigorously characterized over the course of the urea-induced unfolding of the protein chymotrypsin inhibitor 2. By enumerating full entropy production probability distributions as a function of time, we first illustrate that distinct passive and dissipative regimes are present in the denaturation dynamics. Within the dissipative dynamical region, we next find that chymotrypsin inhibitor 2 is strongly driven into unfolded states in which the protein's hydrophobic core has been penetrated by urea molecules and disintegrated. Detailed analyses reveal that urea's interruption of key hydrophobic contacts between core residues causes many of the protein's native structural features to dissolve.

Heat dissipation guides activation in signaling proteins.

Life is fundamentally a nonequilibrium phenomenon. At the expense of dissipated energy, living things perform irreversible processes that allow them to propagate and reproduce. Within cells, evolution has designed nanoscale machines to do meaningful work with energy harnessed from a continuous flux of heat and particles. As dictated by the Second Law of Thermodynamics and its fluctuation theorem corollaries, irreversibility in nonequilibrium processes can be quantified in terms of how much entropy such dynamics produce. In this work, we seek to address a fundamental question linking biology and nonequilibrium physics: can the evolved dissipative pathways that facilitate biomolecular function be identified by their extent of entropy production in general relaxation processes? We here synthesize massive molecular dynamics simulations, Markov state models (MSMs), and nonequilibrium statistical mechanical theory to probe dissipation in two key classes of signaling proteins: kinases and G-protein-coupled receptors (GPCRs). Applying machinery from large deviation theory, we use MSMs constructed from protein simulations to generate dynamics conforming to positive levels of entropy production. We note the emergence of an array of peaks in the dynamical response (transient analogs of phase transitions) that draw the proteins between distinct levels of dissipation, and we see that the binding of ATP and agonist molecules modifies the observed dissipative landscapes. Overall, we find that dissipation is tightly coupled to activation in these signaling systems: dominant entropy-producing trajectories become localized near important barriers along known biological activation pathways. We go on to classify an array of equilibrium and nonequilibrium molecular switches that harmonize to promote functional dynamics.

Markov state models provide insights into dynamic modulation of protein function.

CONSPECTUS: Protein function is inextricably linked to protein dynamics. As we move from a static structural picture to a dynamic ensemble view of protein structure and function, novel computational paradigms are required for observing and understanding conformational dynamics of proteins and its functional implications. In principle, molecular dynamics simulations can provide the time evolution of atomistic models of proteins, but the long time scales associated with functional dynamics make it difficult to observe rare dynamical transitions. The issue of extracting essential functional components of protein dynamics from noisy simulation data presents another set of challenges in obtaining an unbiased understanding of protein motions. Therefore, a methodology that provides a statistical framework for efficient sampling and a human-readable view of the key aspects of functional dynamics from data analysis is required. The Markov state model (MSM), which has recently become popular worldwide for studying protein dynamics, is an example of such a framework. In this Account, we review the use of Markov state models for efficient sampling of the hierarchy of time scales associated with protein dynamics, automatic identification of key conformational states, and the degrees of freedom associated with slow dynamical processes. Applications of MSMs for studying long time scale phenomena such as activation mechanisms of cellular signaling proteins has yielded novel insights into protein function. In particular, from MSMs built using large-scale simulations of GPCRs and kinases, we have shown that complex conformational changes in proteins can be described in terms of structural changes in key structural motifs or "molecular switches" within the protein, the transitions between functionally active and inactive states of proteins proceed via multiple pathways, and ligand or substrate binding modulates the flux through these pathways. Finally, MSMs also provide a theoretical toolbox for studying the effect of nonequilibrium perturbations on conformational dynamics. Considering that protein dynamics in vivo occur under nonequilibrium conditions, MSMs coupled with nonequilibrium statistical mechanics provide a way to connect cellular components to their functional environments. Nonequilibrium perturbations of protein folding MSMs reveal the presence of dynamically frozen glass-like states in their conformational landscape. These frozen states are also observed to be rich in ß-sheets, which indicates their possible role in the nucleation of ß-sheet rich aggregates such as those observed in amyloid-fibril formation. Finally, we describe how MSMs have been used to understand the dynamical behavior of intrinsically disordered proteins such as amyloid-ß, human islet amyloid polypeptide, and p53. While certainly not a panacea for studying functional dynamics, MSMs provide a rigorous theoretical foundation for understanding complex entropically dominated processes and a convenient lens for viewing protein motions.

A Peptide-Coated Gold Nanocluster Exhibits Unique Behavior in Protein Activity Inhibition.

Gold nanoclusters (AuNCs) can be primed for biomedical applications through functionalization with peptide coatings. Often anchored by thiol groups, such peptide coronae not only serve as passivators but can also endow AuNCs with additional bioactive properties. In this work, we use molecular dynamics simulations to study the structure of a tridecapeptide-coated Au25 cluster and its subsequent interactions with the enzyme thioredoxin reductase 1, TrxR1. We find that, in isolation, both the distribution and conformation of the coating peptides fluctuate considerably. When the coated AuNC is placed around TrxR1, however, the motion of the highly charged peptide coating (+5e/peptide) is quickly biased by electrostatic attraction to the protein; the asymmetric coating acts to guide the nanocluster's diffusion toward the enzyme's negatively charged active site. After the AuNC comes into contact with TrxR1, its peptide corona spreads over the protein surface to facilitate stable binding with protein. Though individual salt bridge interactions between the tridecapeptides and TrxR1 are transient in nature, the cooperative binding of the peptide-coated AuNC is very stable, overall. Interestingly, the biased corona peptide motion, the spreading and the cooperation between peptide extensions observed in AuNC binding are reminiscent of bacterial stimulus-driven approaching and adhesion mechanisms mediated by cilia. The prevailing AuNC binding mode we characterize also satisfies a notable hydrophobic interaction seen in the association of thioredoxin to TrxR1, providing a possible explanation for the AuNC binding specificity observed in experiments. Our simulations thus suggest this peptide-coated AuNC serves as an adept thioredoxin mimic that extends an array of auxiliary structural components capable of enhancing interactions with the target protein in question.

Percolation-like phase transitions in network models of protein dynamics.

In broad terms, percolation theory describes the conditions under which clusters of nodes are fully connected in a random network. A percolation phase transition occurs when, as edges are added to a network, its largest connected cluster abruptly jumps from insignificance to complete dominance. In this article, we apply percolation theory to meticulously constructed networks of protein folding dynamics called Markov state models. As rare fluctuations are systematically repressed (or reintroduced), we observe percolation-like phase transitions in protein folding networks: whole sets of conformational states switch from nearly complete isolation to complete connectivity in a rapid fashion. We analyze the general and critical properties of these phase transitions in seven protein systems and discuss how closely dynamics on protein folding landscapes relate to percolation on random lattices.

Dynamical phase transitions reveal amyloid-like states on protein folding landscapes.

Developing an understanding of protein misfolding processes presents a crucial challenge for unlocking the mysteries of human disease. In this article, we present our observations of ß-sheet-rich misfolded states on a number of protein dynamical landscapes investigated through molecular dynamics simulation and Markov state models. We employ a nonequilibrium statistical mechanical theory to identify the glassy states in a protein's dynamics, and we discuss the nonnative, ß-sheet-rich states that play a distinct role in the slowest dynamics within seven protein folding systems. We highlight the fundamental similarity between these states and the amyloid structures responsible for many neurodegenerative diseases, and we discuss potential consequences for mechanisms of protein aggregation and intermolecular amyloid formation.

Emergence of glass-like behavior in Markov state models of protein folding dynamics.

The extent to which glass-like kinetics govern dynamics in protein folding has been heavily debated. Here, we address the subject with an application of space-time perturbation theory to the dynamics of protein folding Markov state models. Borrowing techniques from the s-ensemble method, we argue that distinct active and inactive phases exist for protein folding dynamics, and that kinetics for specific systems can fall into either dynamical regime. We do not, however, observe a true glass transition in any system studied. We go on to discuss how these inactive and active phases might relate to general protein folding properties.

Inclusion of persistence length-based secondary structure in replica field theoretic models of heteropolymer freezing.

The protein folding problem has long represented a "holy grail" in statistical physics due to its physical complexity and its relevance to many human diseases. While past theoretical work has yielded apt descriptions of protein folding landscapes, recent large-scale simulations have provided insights into protein folding that were impractical to obtain from early theories. In particular, the role that non-native contacts play in protein folding, and their relation to the existence of misfolded, ß-sheet rich trap states on folding landscapes, has emerged as a topic of interest in the field. In this paper, we present a modified model of heteropolymer freezing that includes explicit secondary structural characteristics which allow observations of "intramolecular amyloid" states to be probed from a theoretical perspective. We introduce a variable persistence length-based energy penalty to a model Hamiltonian, and we illustrate how this modification alters the phase transitions present in the theory. We find, in particular, that inclusion of this variable persistence length increases both generic freezing and folding temperatures in the model, allowing both folding and glass transitions to occur in a more highly optimized fashion. We go on to discuss how these changes might relate to protein evolution, misfolding, and the emergence of intramolecular amyloid states.

Functional understanding of solvent structure in GroEL cavity through dipole field analysis.

Solvent plays a ubiquitous role in all biophysical phenomena. Yet, just how the molecular nature of water impacts processes in biology remains an important question. While one can simulate the behavior of water near biomolecules such as proteins, it is challenging to gauge the potential structural role solvent plays in mediating both kinetic and equilibrium processes. Here, we propose an analysis scheme for understanding the nature of solvent structure at a local level. We first calculate coarse-grained dipole vector fields for an explicitly solvated system simulated through molecular dynamics. We then analyze correlations between these vector fields to characterize water structure under biologically relevant conditions. In applying our method to the interior of the wild type chaperonin complex GroEL+ES, along with nine additional mutant GroEL complexes, we find that dipole field correlations are strongly related to chaperonin function.

Derivation and assessment of phase-shifted, disordered vector field models for frustrated solvent interactions.

The structure and properties of water at biological interfaces differ drastically from bulk due to effects including confinement and the presence of complicated charge distributions. This non-bulk-like behavior generally arises from water frustration, wherein all favorable interactions among water molecules cannot be simultaneously satisfied. While the frustration of interfacial water is ubiquitous in the cell, the role this frustration plays in mediating biophysical processes like protein folding is not well understood. To investigate the impact of frustration at interfaces, we here derive a general field theoretic model for the interaction of bulk and disordered vector fields at an embedded surface. We calculate thermodynamic and correlation functions for the model in two and three dimensions, and we compare our results to Monte Carlo simulations of lattice system analogs. In our analysis, we see that field-field cross correlations near the interface in the model give rise to a loss in entropy like that seen in glassy systems. We conclude by assessing our theory's utility as a coarse-grained model for water at polar biological interfaces.

Molecular basis of differential HLA class I-restricted T cell recognition of a highly networked HIV peptide.

Cytotoxic-T-lymphocyte (CTL) mediated control of HIV-1 is enhanced by targeting highly networked epitopes in complex with human-leukocyte-antigen-class-I (HLA-I). However, the extent to which the presenting HLA allele contributes to this process is unknown. Here we examine the CTL response to QW9, a highly networked epitope presented by the disease-protective HLA-B57 and disease-neutral HLA-B53. Despite robust targeting of QW9 in persons expressing either allele, T cell receptor (TCR) cross-recognition of the naturally occurring variant QW9_S3T is consistently reduced when presented by HLA-B53 but not by HLA-B57. Crystal structures show substantial conformational changes from QW9-HLA to QW9_S3T-HLA by both alleles. The TCR-QW9-B53 ternary complex structure manifests how the QW9-B53 can elicit effective CTLs and suggests sterically hindered cross-recognition by QW9_S3T-B53. We observe populations of cross-reactive TCRs for B57, but not B53 and also find greater peptide-HLA stability for B57 in comparison to B53. These data demonstrate differential impacts of HLAs on TCR cross-recognition and antigen presentation of a naturally arising variant, with important implications for vaccine design.

Protein folding is mechanistically robust.

Markov state models (MSMs) have proven to be useful tools in simulating large and slowly-relaxing biological systems like proteins. MSMs model proteins through dynamics on a discrete-state energy landscape, allowing molecules to effectively sample large regions of phase space. In this work, we use aspects of MSMs to ask: is protein folding mechanistically robust? We first provide a definition of mechanism in the context of Markovian models, and we later use perturbation theory and the concept of parametric sloppiness to show that parts of the MSM eigenspectrum are resistant to perturbation. We introduce a new, to our knowledge, Bayesian metric by which eigenspectrum robustness can be evaluated, and we discuss the implications of mechanistic robustness and possible new applications of MSMs to understanding biophysical phenomena.

HIV-1 induced changes in HLA-C*03 :  04-presented peptide repertoires lead to reduced engagement of inhibitory natural killer cell receptors.

OBJECTIVE: Viral infections influence intracellular peptide repertoires available for presentation by HLA-I. Alterations in HLA-I/peptide complexes can modulate binding of killer immunoglobuline-like receptors (KIRs) and thereby the function of natural killer (NK) cells. Although multiple studies have provided evidence that HLA-I/KIR interactions play a role in HIV-1 disease progression, the consequence of HIV-1 infection for HLA-I/KIR interactions remain largely unknown. DESIGN: We determined changes in HLA-I presented peptides resulting from HIV-1-infection of primary human CD4 T cells and assessed the impact of changes in peptide repertoires on HLA-I/KIR interactions. METHODS: Liquid chromatography-coupled tandem mass spectrometry to identify HLA-I presented peptides, cell-based in-vitro assays to evaluate functional consequences of alterations in immunopeptidome and atomistic molecular dynamics simulations to confirm experimental data. RESULTS: A total of 583 peptides exclusively presented on HIV-1-infected cells were identified, of which only 0.2% represented HIV-1 derived peptides. Focusing on HLA-C*03 : 04/KIR2DL3 interactions, we observed that HLA-C*03 : 04-presented peptides derived from noninfected CD4 T cells mediated stronger binding of inhibitory KIR2DL3 than peptides derived from HIV-1-infected cells. Furthermore, the most abundant peptide presented by HLA-C*03 : 04 on noninfected CD4 T cells (VIYPARISL) mediated the strongest KIR2DL3-binding, while the most abundant peptide presented on HIV-1-infected cells (YAIQATETL) did not mediate KIR2DL3-binding. Molecular dynamics simulations of HLA-C*03 : 04/KIR2DL3 interactions in the context of these two peptides revealed that VIYPARISL significantly enhanced the HLA-C*03 : 04/peptide contact area to KIR2DL3 compared with YAIQATETL. CONCLUSION: These data demonstrate that HIV-1 infection-induced changes in HLA-I-presented peptides can reduce engagement of inhibitory KIRs, providing a mechanism for enhanced activation of NK cells by virus-infected cells.

Molecular mechanism of Gd@C82(OH)22 increasing collagen expression: Implication for encaging tumor.

Gadolinium-containing fullerenol Gd@C82(OH)22 has demonstrated low-toxicity and highly therapeutic efficacy in inhibiting tumor growth and metastasis through new strategy of encaging cancer, however, little is known about the mechanisms how this nanoparticle regulates fibroblast cells to prison (instead of poison) cancer cells. Here, we report that Gd@C82(OH)22 promote the binding activity of tumor necrosis factor (TNFα) to tumor necrosis factor receptors 2 (TNFR2), activate TNFR2/p38 MAPK signaling pathway to increase cellular collagen expression in fibrosarcoma cells and human primary lung cancer associated fibroblasts isolated from patients. We also employ molecular dynamics simulations to study the atomic-scale mechanisms that dictate how Gd@C82(OH)22 mediates interactions between TNFα and TNFRs. Our data suggest that Gd@C82(OH)22 might enhance the association between TNFα and TNFR2 through a "bridge-like" mode of interaction; by contrast, the fullerenol appears to inhibit TNFα-TNFR1 association by binding to two of the receptor's cysteine-rich domains. In concert, our results uncover a sequential, systemic process by which Gd@C82(OH)22 acts to prison tumor cells, providing new insights into principles of designs of cancer therapeutics.