Platelet Biorheology and Mechanobiology in Thrombosis and Hemostasis: Perspectives from Multiscale Computation.

Thrombosis is the pathological clot formation under abnormal hemodynamic conditions, which can result in vascular obstruction, causing ischemic strokes and myocardial infarction. Thrombus growth under moderate to low shear (<1000 s-1) relies on platelet activation and coagulation. Thrombosis at elevated high shear rates (>10,000 s-1) is predominantly driven by unactivated platelet binding and aggregating mediated by von Willebrand factor (VWF), while platelet activation and coagulation are secondary in supporting and reinforcing the thrombus. Given the molecular and cellular level information it can access, multiscale computational modeling informed by biology can provide new pathophysiological mechanisms that are otherwise not accessible experimentally, holding promise for novel first-principle-based therapeutics. In this review, we summarize the key aspects of platelet biorheology and mechanobiology, focusing on the molecular and cellular scale events and how they build up to thrombosis through platelet adhesion and aggregation in the presence or absence of platelet activation. In particular, we highlight recent advancements in multiscale modeling of platelet biorheology and mechanobiology and how they can lead to the better prediction and quantification of thrombus formation, exemplifying the exciting paradigm of digital medicine.

Exploring Robust Features for Improving Adversarial Robustness.

While deep neural networks (DNNs) have revolutionized many fields, their fragility to carefully designed adversarial attacks impedes the usage of DNNs in safety-critical applications. In this article, we strive to explore the robust features that are not affected by the adversarial perturbations, that is, invariant to the clean image and its adversarial examples (AEs), to improve the model's adversarial robustness. Specifically, we propose a feature disentanglement model to segregate the robust features from nonrobust features and domain-specific features. The extensive experiments on five widely used datasets with different attacks demonstrate that robust features obtained from our model improve the model's adversarial robustness compared to the state-of-the-art approaches. Moreover, the trained domain discriminator is able to identify the domain-specific features from the clean images and AEs almost perfectly. This enables AE detection without incurring additional computational costs. With that, we can also specify different classifiers for clean images and AEs, thereby avoiding any drop in clean image accuracy.

A Multiscale Model for Shear-Mediated Platelet Adhesion Dynamics: Correlating In Silico with In Vitro Results.

Platelet adhesion to blood vessel walls is a key initial event in thrombus formation in both vascular disease processes and prosthetic cardiovascular devices. We extended a deformable multiscale model (MSM) of flowing platelets, incorporating Dissipative Particle Dynamics (DPD) and Coarse-Grained Molecular Dynamics (CGMD) describing molecular-scale intraplatelet constituents and their interaction with surrounding flow, to predict platelet adhesion dynamics under physiological flow shear stresses. Binding of platelet glycoprotein receptor Ibα (GPIbα) to von Willebrand factor (vWF) on the blood vessel wall was modeled by a molecular-level hybrid force field and validated with in vitro microchannel experiments of flowing platelets at 30 dyne/cm2. High frame rate videos of flipping platelets were analyzed with a Semi-Unsupervised Learning System (SULS) machine learning-guided imaging approach to segment platelet geometries and quantify adhesion dynamics parameters. In silico flipping dynamics followed in vitro measurements at 15 and 45 dyne/cm2 with high fidelity, predicting GPIbα-vWF bonding and debonding processes, distribution of bonds strength, and providing a biomechanical insight into initiation of the complex platelet adhesion process. The adhesion model and simulation framework can be further integrated with our established MSMs of platelet activation and aggregation to simulate initial mural thrombus formation on blood vessel walls.

Modeling of the thermal properties of SARS-CoV-2 S-protein.

We calculate the thermal and conformational states of the spike glycoprotein (S-protein) of SARS-CoV-2 at seven temperatures ranging from 3°C to 95°C by all-atom molecular dynamics (MD) µs-scale simulations with the objectives to understand the structural variations on the temperatures and to determine the potential phase transition while trying to correlate such findings of the S-protein with the observed properties of the SARS-CoV2. Our simulations revealed the following thermal properties of the S-protein: 1) It is structurally stable at 3°C, agreeing with observations that the virus stays active for more than two weeks in the cold supply chain; 2) Its structure varies more significantly at temperature values of 60°C-80°C; 3) The sharpest structural variations occur near 60°C, signaling a plausible critical temperature nearby; 4) The maximum deviation of the receptor-binding domain at 37°C, corroborating the anecdotal observations that the virus is most infective at 37°C; 5) The in silico data agree with reported experiments of the SARS-CoV-2 survival times from weeks to seconds by our clustering approach analysis. Our MD simulations at µs scales demonstrated the S-protein's thermodynamics of the critical states at around 60°C, and the stable and denatured states for temperatures below and above this value, respectively.

In Vitro Measurements of Shear-Mediated Platelet Adhesion Kinematics as Analyzed through Machine Learning.

Platelet adhesion to blood vessel walls in shear flow is essential to initiating the blood coagulation cascade and prompting clot formation in vascular disease processes and prosthetic cardiovascular devices. Validation of predictive adhesion kinematics models at the single platelet level is difficult due to gaps in high resolution, dynamic morphological data or a mismatch between simulation and experimental parameters. Gel-filtered platelets were perfused at 30 dyne/cm2 in von Willebrand Factor (vWF)-coated microchannels, with flipping platelets imaged at high spatial and temporal resolution. A semi-unsupervised learning system (SULS), consisting of a series of convolutional neural networks, was used to segment platelet geometry, which was compared with expert-analyzed images. Resulting time-dependent rotational angles were smoothed with wavelet-denoising and shifting techniques to characterize the rotational period and quantify flipping kinematics. We observed that flipping platelets do not follow the previously-modeled modified Jefferey orbit, but are characterized by a longer lift-off and shorter reattachment period. At the juncture of the two periods, rotational velocity approached 257.48 ± 13.31 rad/s. Our SULS approach accurately segmented large numbers of moving platelet images to identify distinct adhesive kinematic characteristics which may further validate the physical accuracy of individual platelet motion in multiscale models of shear-mediated thrombosis.

Rapid analysis of streaming platelet images by semi-unsupervised learning.

We developed a fast and accurate deep learning approach employing a semi-unsupervised learning system (SULS) for capturing the real-time noisy, sparse, and ambiguous images of platelet activation. Outperforming several leading supervised learning methods when applied to segment various platelet morphologies, the SULS detects their complex boundaries at submicron resolutions and it massively decreases to only a few hours for segmenting streaming images of 45 million platelets that would have taken 40 years to annotate manually. For the first time, the fast dynamics of pseudopod formation and platelet morphological changes including membrane tethers and transient tethering to vessels are accurately captured.

A multiscale model for multiple platelet aggregation in shear flow.

We developed a multiscale model for simulating aggregation of multiple, free-flowing platelets in low-intermediate shear viscous flow, in which aggregation is mediated by the interaction of αIIbß3 receptors on the platelet membrane and fibrinogen (Fg). This multiscale model uses coarse grained molecular dynamics (CGMD) for platelets at the microscales and dissipative particle dynamics (DPD) for the shear flow at the macroscales, employing our hybrid aggregation force field for modeling molecular level receptor ligand bonds. We define an aggregation tensor and use it to quantify the molecular level contact characteristics between platelets in an aggregate. We perform numerical studies under different flow conditions for platelet doublets and triplets and evaluate the contact area, detaching force and minimum distance between different pairs of platelets in an aggregate. We also present the dynamics of applied stress and velocity magnitude distributions on the platelet membrane during aggregation and quantify the increase in stress in the contact region under different flow conditions. Integrating the knowledge from our previously validated models, together with new aggregation scenarios, our model can dynamically quantify aggregation characteristics and map stress and velocity distribution on the platelet membrane which are difficult to measure in vitro, thus providing an insight into mechanotransduction bond formation response of platelets to flow-induced shear stresses. This modeling framework, together with the tensor method for quantifying inter-platelet contact, can be extended to simulate and analyze larger aggregates and their adhesive properties.

Supervised machine learning approach to molecular dynamics forecast of SARS-CoV-2 spike glycoproteins at varying temperatures.

ABSTRACT: Molecular dynamics (MD) simulations are a widely used technique in modeling complex nanoscale interactions of atoms and molecules. These simulations can provide detailed insight into how molecules behave under certain environmental conditions. This work explores a machine learning (ML) solution to predicting long-term properties of SARS-CoV-2 spike glycoproteins (S-protein) through the analysis of its nanosecond backbone RMSD (root-mean-square deviation) MD simulation data at varying temperatures. The simulation data were denoised with fast Fourier transforms. The performance of the models was measured by evaluating their mean squared error (MSE) accuracy scores in recurrent forecasts for long-term predictions. The models evaluated include k-nearest neighbors (kNN) regression models, as well as GRU (gated recurrent unit) neural networks and LSTM (long short-term memory) autoencoder models. Results demonstrated that the kNN model achieved the greatest accuracy in forecasts with MSE scores over around 0.01 nm less than those of the GRU model and the LSTM autoencoder. Furthermore, it demonstrated that the kNN model accuracy increases with data size but can still forecast relatively well when trained on small amounts of data, having achieved MSE scores of around 0.02 nm when trained on 10,000 ns of simulation data. This study provides valuable information on the feasibility of accelerating the MD simulation process through training and predicting supervised ML models, which is particularly applicable in time-sensitive studies. GRAPHIC ABSTRACT: SARS-CoV-2 spike glycoprotein molecular dynamics simulation. Extraction and denoising of backbone RMSD data. Evaluation of k-nearest neighbors regression, GRU neural network, and LSTM autoencoder models in recurrent forecasting for long-term property predictions.

A predictive multiscale model for simulating flow-induced platelet activation: Correlating in silico results with in vitro results.

Flow-induced platelet activation prompts complex filopodial formation. Continuum methods fail to capture such molecular-scale mechanisms. A multiscale numerical model was developed to simulate this activation process, where a Dissipative Particle Dynamics (DPD) model of viscous blood flow is interfaced with a Coarse Grained Molecular Dynamics (CGMD) platelet model. Embedded in DPD blood flow, the macroscopic dynamic stresses are interactively transferred to the CGMD model, inducing intra-platelet associated events. The platelets activate by a biomechanical transductive linkage chain and dynamically change their shape in response. The models are fully coupled via a hybrid-potential interface and multiple time-stepping (MTS) schemes for handling the disparity between the spatiotemporal scales. Cumulative hemodynamic stresses that may lead to platelet activation are mapped on the surface membrane and simultaneously transmitted to the cytoplasm and cytoskeleton. Upon activation, the flowing platelets lose their quiescent discoid shape and evolve by forming filopodia. The model predictions were validated by a set of in vitro experiments, Platelets were exposed to various combinations of shear stresses and durations in our programmable hemodynamic shearing device (HSD). Their shape change was measured at multiple time points using scanning electron microscopy (SEM). The CGMD model parameters were fine-tuned by interrogating a parameter space established in these experiments. Segmentation of the SEM imaging streams was conducted by a deep machine learning system. This model can be further employed to simulate shear mediated platelet activation thrombosis initiation and to study the effects of modulating platelet properties to enhance their shear resistance via mechanotransduction pathways.

Artificial Intelligence for Accelerating Time Integrations in Multiscale Modeling.

We developed a novel data-driven Artificial Intelligence-enhanced Adaptive Time Stepping algorithm (AI-ATS) that can adapt timestep sizes to underlying biophysical dynamics. We demonstrated its values in solving a complex biophysical problem, at multiple spatiotemporal scales, that describes platelet dynamics in shear blood flow. In order to achieve a significant speedup of this computationally demanding problem, we integrated a framework of novel AI algorithms into the solution of the platelet dynamics equations. Our framework involves recurrent neural network-based autoencoders by the Long Short-Term Memory and the Gated Recurrent Units as the first step for memorizing the dynamic states in long-term dependencies for the input time series, followed by two fully-connected neural networks to optimize timestep sizes and step jumps. The computational efficiency of our AI-ATS is underscored by assessing the accuracy and speed of a multiscale simulation of the platelet with the standard time stepping algorithm (STS). By adapting the timestep size, our AI-ATS guides the omission of multiple redundant time steps without sacrificing significant accuracy of the dynamics. Compared to the STS, our AI-ATS achieved a reduction of 40% unnecessary calculations while bounding the errors of mechanical and thermodynamic properties to 3%.

Online Machine Learning for Accelerating Molecular Dynamics Modeling of Cells.

We developed a biomechanics-informed online learning framework to learn the dynamics with ground truth generated with multiscale modeling simulation. It was built on Summit-like supercomputers, which were also used to benchmark and validate our framework on one physiologically significant modeling of deformable biological cells. We generalized the century-old equation of Jeffery orbits to a new equation of motion with additional parameters to account for the flow conditions and the cell deformability. Using simulation data at particle-based resolutions for flowing cells and the learned parameters from our framework, we validated the new equation by the motions, mostly rotations, of a human platelet in shear blood flow at various shear stresses and platelet deformability. Our online framework, which surrogates redundant computations in the conventional multiscale modeling by solutions of our learned equation, accelerates the conventional modeling by three orders of magnitude without visible loss of accuracy.

A Multiscale Model for Recruitment Aggregation of Platelets by Correlating with In Vitro Results.

INTRODUCTION: We developed a multiscale model to simulate the dynamics of platelet aggregation by recruitment of unactivated platelets flowing in viscous shear flows by an activated platelet deposited onto a blood vessel wall. This model uses coarse grained molecular dynamics (CGMD) for platelets at the microscale and dissipative particle dynamics (DPD) for the shear flow at the macroscale. Under conditions of relatively low shear, aggregation is mediated by fibrinogen via αIIbß3 receptors. METHODS: The binding of αIIbß3 and fibrinogen is modeled by a molecular-level hybrid force field consisting of Morse potential and Hooke law for the nonbonded and bonded interactions, respectively. The force field, parametrized in two different interaction scales, is calculated by correlating with the platelet contact area measured in vitro and the detaching force between αIIbß3 and fibrinogen. RESULTS: Using our model, we derived, the relationship between recruitment force and distance between the centers of mass of two platelets, by integrating the molecular-scale inter-platelet interactions during recruitment aggregation in shear flows. Our model indicates that assuming a rigid-platelet model, underestimates the contact area by 89% and the detaching force by 93% as compared to a model that takes into account the platelet deformability leading to a prediction of a significantly lower attachment during recruitment. CONCLUSIONS: The molecular-level predictive capability of our model sheds a light on differences observed between transient and permanent platelet aggregation patterns. The model and simulation framework can be further adapted to simulate initial thrombus formation involving multiple flowing platelets as well as deposition and adhesion onto blood vessels.

A multiscale biomechanical model of platelets: Correlating with in-vitro results.

Using dissipative particle dynamics (DPD) combined with coarse grained molecular dynamics (CGMD) approaches, we developed a multiscale deformable platelet model to accurately describe the molecular-scale intra-platelet constituents and biomechanical properties of platelets in blood flow. Our model includes the platelet bilayer membrane, cytoplasm and an elaborate elastic cytoskeleton. Correlating numerical simulations with published in-vitro experiments, we validated the biorheology of the cytoplasm, the elastic response of membrane to external stresses, and the stiffness of the cytoskeleton actin filaments, resulting in an accurate representation of the molecular-level biomechanical microstructures of platelets. This enabled us to study the mechanotransduction process of the hemodynamic stresses acting onto the platelet membrane and transmitted to these intracellular constituents. The platelets constituents continuously deform in response to the flow induced stresses. To the best of our knowledge, this is the first molecular-scale platelet model that can be used to accurately predict platelets activation mechanism leading to thrombus formation in prosthetic cardiovascular devices and in vascular disease processes. This model can be further employed to study the effects of novel therapeutic approaches of modulating platelet properties to enhance their shear resistance via mechanotransduction pathways.

Scalability Test of Multiscale Fluid-Platelet Model for Three Top Supercomputers.

We have tested the scalability of three supercomputers: the Tianhe-2, Stampede and CS-Storm with multiscale fluid-platelet simulations, in which a highly-resolved and efficient numerical model for nanoscale biophysics of platelets in microscale viscous biofluids is considered. Three experiments involving varying problem sizes were performed: Exp-S: 680,718-particle single-platelet; Exp-M: 2,722,872-particle 4-platelet; and Exp-L: 10,891,488-particle 16-platelet. Our implementations of multiple time-stepping (MTS) algorithm improved the performance of single time-stepping (STS) in all experiments. Using MTS, our model achieved the following simulation rates: 12.5, 25.0, 35.5 µs/day for Exp-S and 9.09, 6.25, 14.29 µs/day for Exp-M on Tianhe-2, CS-Storm 16-K80 and Stampede K20. The best rate for Exp-L was 6.25 µs/day for Stampede. Utilizing current advanced HPC resources, the simulation rates achieved by our algorithms bring within reach performing complex multiscale simulations for solving vexing problems at the interface of biology and engineering, such as thrombosis in blood flow which combines millisecond-scale hematology with microscale blood flow at resolutions of micro-to-nanoscale cellular components of platelets. This study of testing the performance characteristics of supercomputers with advanced computational algorithms that offer optimal trade-off to achieve enhanced computational performance serves to demonstrate that such simulations are feasible with currently available HPC resources.

A Multiple Time Stepping Algorithm for Efficient Multiscale Modeling of Platelets Flowing in Blood Plasma.

We developed a multiple time-stepping (MTS) algorithm for multiscale modeling of the dynamics of platelets flowing in viscous blood plasma. This MTS algorithm improves considerably the computational efficiency without significant loss of accuracy. This study of the dynamic properties of flowing platelets employs a combination of the dissipative particle dynamics (DPD) and the coarse-grained molecular dynamics (CGMD) methods to describe the dynamic microstructures of deformable platelets in response to extracellular flow-induced stresses. The disparate spatial scales between the two methods are handled by a hybrid force field interface. However, the disparity in temporal scales between the DPD and CGMD that requires time stepping at microseconds and nanoseconds respectively, represents a computational challenge that may become prohibitive. Classical MTS algorithms manage to improve computing efficiency by multi-stepping within DPD or CGMD for up to one order of magnitude of scale differential. In order to handle 3-4 orders of magnitude disparity in the temporal scales between DPD and CGMD, we introduce a new MTS scheme hybridizing DPD and CGMD by utilizing four different time stepping sizes. We advance the fluid system at the largest time step, the fluid-platelet interface at a middle timestep size, and the nonbonded and bonded potentials of the platelet structural system at two smallest timestep sizes. Additionally, we introduce parameters to study the relationship of accuracy versus computational complexities. The numerical experiments demonstrated 3000x reduction in computing time over standard MTS methods for solving the multiscale model. This MTS algorithm establishes a computationally feasible approach for solving a particle-based system at multiple scales for performing efficient multiscale simulations.

A phenomenological particle-based platelet model for simulating filopodia formation during early activation.

We developed a phenomenological three-dimensional platelet model to characterize the filopodia formation observed during early stage platelet activation. Departing from continuum mechanics based approaches, this coarse-grained molecular dynamics (CGMD) particle-based model can deform to emulate the complex shape change and filopodia formation that platelets undergo during activation. The platelet peripheral zone is modeled with a two-layer homogeneous elastic structure represented by spring-connected particles. The structural zone is represented by a cytoskeletal assembly comprising of a filamentous core and filament bundles supporting the platelet's discoid shape, also modeled by spring-connected particles. The interior organelle zone is modeled by homogeneous cytoplasm particles that facilitate the platelet deformation. Nonbonded interactions among the discrete particles of the membrane, the cytoskeletal assembly, and the cytoplasm are described using the Lennard-Jones potential with empirical constants. By exploring the parameter space of this CGMD model, we have successfully simulated the dynamics of varied filopodia formations. Comparative analyses of length and thickness of filopodia show that our numerical simulations are in agreement with experimental measurements of flow-induced activated platelets. Copyright © 2015 John Wiley & Sons, Ltd.

Multiscale Particle-Based Modeling of Flowing Platelets in Blood Plasma Using Dissipative Particle Dynamics and Coarse Grained Molecular Dynamics.

We developed a multiscale particle-based model of platelets, to study the transport dynamics of shear stresses between the surrounding fluid and the platelet membrane. This model facilitates a more accurate prediction of the activation potential of platelets by viscous shear stresses - one of the major mechanisms leading to thrombus formation in cardiovascular diseases and in prosthetic cardiovascular devices. The interface of the model couples coarse-grained molecular dynamics (CGMD) with dissipative particle dynamics (DPD). The CGMD handles individual platelets while the DPD models the macroscopic transport of blood plasma in vessels. A hybrid force field is formulated for establishing a functional interface between the platelet membrane and the surrounding fluid, in which the microstructural changes of platelets may respond to the extracellular viscous shear stresses transferred to them. The interaction between the two systems preserves dynamic properties of the flowing platelets, such as the flipping motion. Using this multiscale particle-based approach, we have further studied the effects of the platelet elastic modulus by comparing the action of the flow-induced shear stresses on rigid and deformable platelet models. The results indicate that neglecting the platelet deformability may overestimate the stress on the platelet membrane, which in turn may lead to erroneous predictions of the platelet activation under viscous shear flow conditions. This particle-based fluid-structure interaction multiscale model offers for the first time a computationally feasible approach for simulating deformable platelets interacting with viscous blood flow, aimed at predicting flow induced platelet activation by using a highly resolved mapping of the stress distribution on the platelet membrane under dynamic flow conditions.

Parameterizing the Morse Potential for Coarse-Grained Modeling of Blood Plasma.

Multiscale simulations of fluids such as blood represent a major computational challenge of coupling the disparate spatiotemporal scales between molecular and macroscopic transport phenomena characterizing such complex fluids. In this paper, a coarse-grained (CG) particle model is developed for simulating blood flow by modifying the Morse potential, traditionally used in Molecular Dynamics for modeling vibrating structures. The modified Morse potential is parameterized with effective mass scales for reproducing blood viscous flow properties, including density, pressure, viscosity, compressibility and characteristic flow dynamics of human blood plasma fluid. The parameterization follows a standard inverse-problem approach in which the optimal micro parameters are systematically searched, by gradually decoupling loosely correlated parameter spaces, to match the macro physical quantities of viscous blood flow. The predictions of this particle based multiscale model compare favorably to classic viscous flow solutions such as Counter-Poiseuille and Couette flows. It demonstrates that such coarse grained particle model can be applied to replicate the dynamics of viscous blood flow, with the advantage of bridging the gap between macroscopic flow scales and the cellular scales characterizing blood flow that continuum based models fail to handle adequately.

Viscous flow simulation in a stenosis model using discrete particle dynamics: a comparison between DPD and CFD.

Flow and stresses induced by blood flow acting on the blood cellular constituents can be represented to a certain extent by a continuum mechanics approach down to the order of the µm level. However, the molecular effects of, e.g., adhesion/aggregation bonds of blood clotting can be on the order of nm. The coupling of the disparate length and timescales between such molecular levels and macroscopic transport represents a major computational challenge. To address this challenge, a multiscale numerical approach based on discrete particle dynamics (DPD) methodology derived from molecular dynamics (MD) principles is proposed. The feasibility of the approach was firstly tested for its ability to simulate viscous flow conditions. Simulations were conducted in low Reynolds numbers flows (Re = 25-33) through constricted tubes representing blood vessels with various degrees of stenosis. Multiple discrete particles interacting with each other were simulated, with 1.24-1.36 million particles representing the flow domain and 0.4 million particles representing the vessel wall. The computation was carried out on the massive parallel supercomputer NY BlueGene/L employing NAMD-a parallel MD package for high performance computing (HPC). Typical recirculation zones were formed distal to the stenoses. The velocity profiles and recirculation zones were in excellent agreement with computational fluid dynamics (CFD) 3D Navier-Stokes viscous fluid flow simulations and with classic numerical and experimental results by YC Fung in constricted tubes. This feasibility analysis demonstrates the potential of a methodology that widely departs from a continuum approach to simulate multiscale phenomena such as flow induced blood clotting.