A Data-Driven Dimensionality Reduction Approach to Compare and Classify Lipid Force Fields

Sampling Real-Time Atomic Dynamics in Metal Nanoparticles by Combining Experiments, Simulations, and Machine Learning

Even at low temperatures, metal nanoparticles (NPs) possess atomic dynamics that are key for their properties but challenging to elucidate. Recent experimental advances allow obtaining atomic-resolution snapshots of the NPs in realistic regimes, but data acquisition limitations hinder the experimental reconstruction of the atomic dynamics present within them. Molecular simulations have the advantage that these allow directly tracking the motion of atoms over time. However, these typically start from ideal/perfect NP structures and, suffering from sampling limits, provide results that are often dependent on the initial/putative structure and remain purely indicative. Here, by combining state-of-the-art experimental and computational approaches, how it is possible to tackle the limitations of both approaches and resolve the atomistic dynamics present in metal NPs in realistic conditions is demonstrated. Annular dark-field scanning transmission electron microscopy enables the acquisition of ten high-resolution images of an Au NP at intervals of 0.6 s. These are used to reconstruct atomistic 3D models of the real NP used to run ten independent molecular dynamics simulations. Machine learning analyses of the simulation trajectories allow resolving the real-time atomic dynamics present within the NP. This provides a robust combined experimental/computational approach to characterize the structural dynamics of metal NPs in realistic conditions.

Data-driven discovery of innate immunomodulators via machine learning-guided high throughput screening

The innate immune response is vital for the success of prophylactic vaccines and immunotherapies. Control of signaling in innate immune pathways can improve prophylactic vaccines by inhibiting unfavorable systemic inflammation and immunotherapies by enhancing immune stimulation. In this work, we developed a machine learning-enabled active learning pipeline to guide in vitro experimental screening and discovery of small molecule immunomodulators that improve immune responses by altering the signaling activity of innate immune responses stimulated by traditional pattern recognition receptor agonists. Molecules were tested by in vitro high throughput screening (HTS) where we measured modulation of the nuclear factor κ-light-chain-enhancer of activated B-cells (NF-κB) and the interferon regulatory factors (IRF) pathways. These data were used to train data-driven predictive models linking molecular structure to modulation of the NF-κB and IRF responses using deep representational learning, Gaussian process regression, and Bayesian optimization. By interleaving successive rounds of model training and in vitro HTS, we performed an active learning-guided traversal of a 139 998 molecule library. After sampling only ∼2% of the library, we discovered viable molecules with unprecedented immunomodulatory capacity, including those capable of suppressing NF-κB activity by up to 15-fold, elevating NF-κB activity by up to 5-fold, and elevating IRF activity by up to 6-fold. We extracted chemical design rules identifying particular chemical fragments as principal drivers of specific immunomodulation behaviors. We validated the immunomodulatory effect of a subset of our top candidates by measuring cytokine release profiles. Of these, one molecule induced a 3-fold enhancement in IFN-β production when delivered with a cyclic di-nucleotide stimulator of interferon genes (STING) agonist. In sum, our machine learning-enabled screening approach presents an efficient immunomodulator discovery pipeline that has furnished a library of novel small molecules with a strong capacity to enhance or suppress innate immune signaling pathways to shape and improve prophylactic vaccination and immunotherapies.

The SIRAH force field: A suite for simulations of complex biological systems at the coarse-grained and multiscale levels

The different combinations of molecular dynamics simulations with coarse-grained representations have acquired considerable popularity among the scientific community. Especially in biocomputing, the significant speedup granted by simplified molecular models opened the possibility of increasing the diversity and complexity of macromolecular systems, providing realistic insights on large assemblies for more extended time windows. However, a holistic view of biological ensembles' structural and dynamic features requires a self-consistent force field, namely, a set of equations and parameters that describe the intra and intermolecular interactions among moieties of diverse chemical nature (i.e., nucleic and amino acids, lipids, solvent, ions, etc.). Nevertheless, examples of such force fields are scarce in the literature at the fully atomistic and coarse-grained levels. Moreover, the number of force fields capable of handling simultaneously different scales is restricted to a handful. Among those, the SIRAH force field, developed in our group, furnishes a set of topologies and tools that facilitate the setting up and running of molecular dynamics simulations at the coarse-grained and multiscale levels. SIRAH uses the same classical pairwise Hamiltonian function implemented in the most popular molecular dynamics software. In particular, it runs natively in AMBER and Gromacs engines, and porting it to other simulation packages is straightforward. This review describes the underlying philosophy behind the development of SIRAH over the years and across families of biological molecules, discussing current limitations and future implementations.

Detecting dynamic domains and local fluctuations in complex molecular systems via timelapse neighbors shuffling

It is known that the behavior of many complex systems is controlled by local dynamic rearrangements or fluctuations occurring within them. Complex molecular systems, composed of many molecules interacting with each other in a Brownian storm, make no exception. Despite the rise of machine learning and of sophisticated structural descriptors, detecting local fluctuations and collective transitions in complex dynamic ensembles remains often difficult. Here, we show a machine learning framework based on a descriptor which we name Local Environments and Neighbors Shuffling (LENS), that allows identifying dynamic domains and detecting local fluctuations in a variety of systems in an abstract and efficient way. By tracking how much the microscopic surrounding of each molecular unit changes over time in terms of neighbor individuals, LENS allows characterizing the global (macroscopic) dynamics of molecular systems in phase transition, phases-coexistence, as well as intrinsically characterized by local fluctuations (e.g., defects). Statistical analysis of the LENS time series data extracted from molecular dynamics trajectories of, for example, liquid-like, solid-like, or dynamically diverse complex molecular systems allows tracking in an efficient way the presence of different dynamic domains and of local fluctuations emerging within them. The approach is found robust, versatile, and applicable independently of the features of the system and simply provided that a trajectory containing information on the relative motion of the interacting units is available. We envisage that "such a LENS" will constitute a precious basis for exploring the dynamic complexity of a variety of systems and, given its abstract definition, not necessarily of molecular ones.

Machine learning of atomic dynamics and statistical surface identities in gold nanoparticles

It is known that metal nanoparticles (NPs) may be dynamic and atoms may move within them even at fairly low temperatures. Characterizing such complex dynamics is key for understanding NPs' properties in realistic regimes, but detailed information on, e.g., the stability, survival, and interconversion rates of the atomic environments (AEs) populating them are non-trivial to attain. In this study, we decode the intricate atomic dynamics of metal NPs by using a machine learning approach analyzing high-dimensional data obtained from molecular dynamics simulations. Using different-shape gold NPs as a representative example, an AEs' dictionary allows us to label step-by-step the individual atoms in the NPs, identifying the native and non-native AEs and populating them along the MD simulations at various temperatures. By tracking the emergence, annihilation, lifetime, and dynamic interconversion of the AEs, our approach permits estimating a "statistical equivalent identity" for metal NPs, providing a comprehensive picture of the intrinsic atomic dynamics that shape their properties.

Integrating Machine Learning in the Coarse-Grained Molecular Simulation of Polymers

Machine learning (ML) is having an increasing impact on the physical sciences, engineering, and technology and its integration into molecular simulation frameworks holds great potential to expand their scope of applicability to complex materials and facilitate fundamental knowledge and reliable property predictions, contributing to the development of efficient materials design routes. The application of ML in materials informatics in general, and polymer informatics in particular, has led to interesting results, however great untapped potential lies in the integration of ML techniques into the multiscale molecular simulation methods for the study of macromolecular systems, specifically in the context of Coarse Grained (CG) simulations. In this Perspective, we aim at presenting the pioneering recent research efforts in this direction and discussing how these new ML-based techniques can contribute to critical aspects of the development of multiscale molecular simulation methods for bulk complex chemical systems, especially polymers. Prerequisites for the implementation of such ML-integrated methods and open challenges that need to be met toward the development of general systematic ML-based coarse graining schemes for polymers are discussed.

Unsupervised Data-Driven Reconstruction of Molecular Motifs in Simple to Complex Dynamic Micelles

The reshuffling mobility of molecular building blocks in self-assembled micelles is a key determinant of many their interesting properties, from emerging morphologies and surface compartmentalization, to dynamic reconfigurability and stimuli-responsiveness. However, the microscopic details of such complex structural dynamics are typically nontrivial to elucidate, especially in multicomponent assemblies. Here we show a machine-learning approach that allows us to reconstruct the structural and dynamic complexity of mono- and bicomponent surfactant micelles from high-dimensional data extracted from equilibrium molecular dynamics simulations. Unsupervised clustering of smooth overlap of atomic position (SOAP) data enables us to identify, in a set of multicomponent surfactant micelles, the dominant local molecular environments that emerge within them and to retrace their dynamics, in terms of exchange probabilities and transition pathways of the constituent building blocks. Tested on a variety of micelles differing in size and in the chemical nature of the constitutive self-assembling units, this approach effectively recognizes the molecular motifs populating them in an exquisitely agnostic and unsupervised way, and allows correlating them to their composition in terms of constitutive surfactant species.

DADApy: Distance-based analysis of data-manifolds in Python

DADApy is a Python software package for analyzing and characterizing high-dimensional data manifolds. It provides methods for estimating the intrinsic dimension and the probability density, for performing density-based clustering, and for comparing different distance metrics. We review the main functionalities of the package and exemplify its usage in a synthetic dataset and in a real-world application. DADApy is freely available under the open-source Apache 2.0 license.

Classifying soft self-assembled materials via unsupervised machine learning of defects

Unlike molecular crystals, soft self-assembled fibers, micelles, vesicles, etc., exhibit a certain order in the arrangement of their constitutive monomers but also high structural dynamicity and variability. Defects and disordered local domains that continuously form-and-repair in their structures impart to such materials unique adaptive and dynamical properties, which make them, e.g., capable to communicate with each other. However, objective criteria to compare such complex dynamical features and to classify soft supramolecular materials are non-trivial to attain. Here we show a data-driven workflow allowing us to achieve this goal. Building on unsupervised clustering of Smooth Overlap of Atomic Position (SOAP) data obtained from equilibrium molecular dynamics simulations, we can compare a variety of soft supramolecular assemblies via a robust SOAP metric. This provides us with a data-driven "defectometer" to classify different types of supramolecular materials based on the structural dynamics of the ordered/disordered local molecular environments that statistically emerge within them.

Ephemeral ice-like local environments in classical rigid models of liquid water

Despite great efforts over the past 50 years, the simulation of water still presents significant challenges and open questions. At room temperature and pressure, the collective molecular interactions and dynamics of water molecules may form local structural arrangements that are non-trivial to classify. Here, we employ a data-driven approach built on Smooth Overlap of Atomic Position (SOAP) that allows us to compare and classify how widely used classical models represent liquid water. Macroscopically, the obtained results are rationalized based on water thermodynamic observables. Microscopically, we directly observe how transient ice-like ordered environments may dynamically/statistically form in liquid water, even above freezing temperature, by comparing the SOAP spectra for different ice structures with those of the simulated liquid systems. This confirms recent ab initio-based calculations but also reveals how the emergence of ephemeral local ice-like environments in liquid water at room conditions can be captured by classical water models.

The stressed life of a lipid in the Zika virus membrane

Protein-lipid interactions modulate a plethora of physiopathologic processes and have been the subject of countless studies. However, these kinds of interactions in the context of viral envelopes have remained relatively unexplored, partially because the intrinsically small dimensions of the molecular systems escape to the current resolution of experimental techniques. However, coarse-grained and multiscale simulations may fill that niche, providing nearly atomistic resolution at an affordable computational price. Here we use multiscale simulations to characterize the lipid-protein interactions in the envelope of the Zika Virus, a prominent member of the Flavivirus genus. Comparisons between the viral envelope and simpler molecular systems indicate that the viral membrane is under extreme pressures and asymmetric forces. Furthermore, the dense net of protein-protein contacts established by the envelope proteins creates poorly solvated regions that destabilize the external leaflet leading to a decoupled dynamics between both membrane layers. These findings lead to the idea that the Flaviviral membrane may store a significant amount of elastic energy, playing an active role in the membrane fusion process.

Identifying nonadditive contributions to the hydrophobicity of chemically heterogeneous surfaces via dual-loop active learning

Hydrophobic interactions drive numerous biological and synthetic processes. The materials used in these processes often possess chemically heterogeneous surfaces that are characterized by diverse chemical groups positioned in close proximity at the nanoscale; examples include functionalized nanomaterials and biomolecules, such as proteins and peptides. Nonadditive contributions to the hydrophobicity of such surfaces depend on the chemical identities and spatial patterns of polar and nonpolar groups in ways that remain poorly understood. Here, we develop a dual-loop active learning framework that combines a fast reduced-accuracy method (a convolutional neural network) with a slow higher-accuracy method (molecular dynamics simulations with enhanced sampling) to efficiently predict the hydration free energy, a thermodynamic descriptor of hydrophobicity, for nearly 200 000 chemically heterogeneous self-assembled monolayers (SAMs). Analysis of this dataset reveals that SAMs with distinct polar groups exhibit substantial variations in hydrophobicity as a function of their composition and patterning, but the clustering of nonpolar groups is a common signature of highly hydrophobic patterns. Further molecular dynamics analysis relates such clustering to the perturbation of interfacial water structure. These results provide new insight into the influence of chemical heterogeneity on hydrophobicity via quantitative analysis of a large set of surfaces, enabled by the active learning approach.

Neural Upscaling from Residue-Level Protein Structure Networks to Atomistic Structures

Coarse-graining is a powerful tool for extending the reach of dynamic models of proteins and other biological macromolecules. Topological coarse-graining, in which biomolecules or sets thereof are represented via graph structures, is a particularly useful way of obtaining highly compressed representations of molecular structures, and simulations operating via such representations can achieve substantial computational savings. A drawback of coarse-graining, however, is the loss of atomistic detail—an effect that is especially acute for topological representations such as protein structure networks (PSNs). Here, we introduce an approach based on a combination of machine learning and physically-guided refinement for inferring atomic coordinates from PSNs. This “neural upscaling” procedure exploits the constraints implied by PSNs on possible configurations, as well as differences in the likelihood of observing different configurations with the same PSN. Using a 1 μs atomistic molecular dynamics trajectory of Aβ1–40, we show that neural upscaling is able to effectively recapitulate detailed structural information for intrinsically disordered proteins, being particularly successful in recovering features such as transient secondary structure. These results suggest that scalable network-based models for protein structure and dynamics may be used in settings where atomistic detail is desired, with upscaling employed to impute atomic coordinates from PSNs.

Subdiffusive-Brownian crossover in membrane proteins: a generalized Langevin equation-based approach

In this work, we propose a generalized Langevin equation-based model to describe the lateral diffusion of a protein in a lipid bilayer. The memory kernel is represented in terms of a viscous (instantaneous) and an elastic (noninstantaneous) component modeled through a Dirac δ function and a three-parameter Mittag-Leffler type function, respectively. By imposing a specific relationship between the parameters of the three-parameter Mittag-Leffler function, the different dynamical regimes-namely ballistic, subdiffusive, and Brownian, as well as the crossover from one regime to another-are retrieved. Within this approach, the transition time from the ballistic to the subdiffusive regime and the spectrum of relaxation times underlying the transition from the subdiffusive to the Brownian regime are given. The reliability of the model is tested by comparing the mean-square displacement derived in the framework of this model and the mean-square displacement of a protein diffusing in a membrane calculated through molecular dynamics simulations.

Mechanistic Understanding from Molecular Dynamics in Pharmaceutical Research 2: Lipid Membrane in Drug Design

We review the use of molecular dynamics (MD) simulation as a drug design tool in the context of the role that the lipid membrane can play in drug action, i.e., the interaction between candidate drug molecules and lipid membranes. In the standard "lock and key" paradigm, only the interaction between the drug and a specific active site of a specific protein is considered; the environment in which the drug acts is, from a biophysical perspective, far more complex than this. The possible mechanisms though which a drug can be designed to tinker with physiological processes are significantly broader than merely fitting to a single active site of a single protein. In this paper, we focus on the role of the lipid membrane, arguably the most important element outside the proteins themselves, as a case study. We discuss work that has been carried out, using MD simulation, concerning the transfection of drugs through membranes that act as biological barriers in the path of the drugs, the behavior of drug molecules within membranes, how their collective behavior can affect the structure and properties of the membrane and, finally, the role lipid membranes, to which the vast majority of drug target proteins are associated, can play in mediating the interaction between drug and target protein. This review paper is the second in a two-part series covering MD simulation as a tool in pharmaceutical research; both are designed as pedagogical review papers aimed at both pharmaceutical scientists interested in exploring how the tool of MD simulation can be applied to their research and computational scientists interested in exploring the possibility of a pharmaceutical context for their research.

LiPyphilic: A Python Toolkit for the Analysis of Lipid Membrane Simulations

Molecular dynamics simulations are now widely used to study emergent phenomena in lipid membranes with complex compositions. Here, we present LiPyphilic-a fast, fully tested, and easy-to-install Python package for analyzing such simulations. Analysis tools in LiPyphilic include the identification of cholesterol flip-flop events, the classification of local lipid environments, and the degree of interleaflet registration. LiPyphilic is both force field- and resolution-agnostic, and by using the powerful atom selection language of MDAnalysis, it can handle membranes with highly complex compositions. LiPyphilic also offers two on-the-fly trajectory transformations to (i) fix membranes split across periodic boundaries and (ii) perform nojump coordinate unwrapping. Our implementation of nojump unwrapping accounts for fluctuations in the box volume under the NPT ensemble-an issue that most current implementations have overlooked. The full documentation of LiPyphilic, including installation instructions and links to interactive online tutorials, is available at https://lipyphilic.readthedocs.io/en/latest.

Living supramolecular polymerization of fluorinated cyclohexanes

The development of powerful methods for living covalent polymerization has been a key driver of progress in organic materials science. While there have been remarkable reports on living supramolecular polymerization recently, the scope of monomers is still narrow and a simple solution to the problem is elusive. Here we report a minimalistic molecular platform for living supramolecular polymerization that is based on the unique structure of all-cis 1,2,3,4,5,6-hexafluorocyclohexane, the most polar aliphatic compound reported to date. We use this large dipole moment (6.2 Debye) not only to thermodynamically drive the self-assembly of supramolecular polymers, but also to generate kinetically trapped monomeric states. Upon addition of well-defined seeds, we observed that the dormant monomers engage in a kinetically controlled supramolecular polymerization. The obtained nanofibers have an unusual double helical structure and their length can be controlled by the ratio between seeds and monomers. The successful preparation of supramolecular block copolymers demonstrates the versatility of the approach.