arXiv subjects

Michele Parrinello

Publications and source records attributed to Michele Parrinello.

At least 37 records · Page 2Linked to original sources

arXiv2024

Transfer learning for atomistic simulations using GNNs and kernel mean embeddings

Interatomic potentials learned using machine learning methods have been successfully applied to atomistic simulations. However, accurate models require large training datasets, while generating reference calculations is computationally demanding. To bypass this difficulty, we propose a transfer learning algorithm that leverages the ability of graph neural networks (GNNs) to represent chemical environments together with kernel mean embeddings. We extract a feature map from GNNs pre-trained on the OC20 dataset and use it to learn the potential energy surface from system-specific datasets of catalytic processes. Our method is further enhanced by incorporating into the kernel the chemical species information, resulting in improved performance and interpretability. We test our approach on a series of realistic datasets of increasing complexity, showing excellent generalization and transferability performance, and improving on methods that rely on GNNs or ridge regression alone, as well as similar fine-tuning approaches.

cs.LG↗

arXiv2023

Intramolecular and water mediated tautomerism of solvated glycine

The understanding of prototropic tautomerism in water and the characterization of solvent effects on protomeric equilibrium pose significant challenges. Using molecular dynamics simulations based on state-of-the-art deep learning potential and enhanced sampling methods, we provide a comprehensive description of all configurational transformations in glycine solvated in water and determine accurate free energy profiles of these processes. We observe that the tautomerism between the neutral and zwitterionic forms of solvated glycine can occur by both intramolecular proton transfer in glycine and intermolecular proton transfer in the contact ion pair (anionic glycine and hydronium ion) or the separated ion pair (cationic glycine and hydroxide ion).

cond-mat.soft↗

arXiv2023

Structure and polymerization of liquid sulfur across the $λ$-transition

The anomalous $λ$-transition of liquid sulfur, which is supposed to be related to the transformation of eight-membered sulfur rings into long polymeric chains, has attracted considerable attention. However, a detailed description of the underlying dynamical polymerization process is still missing. Here, we study the structures and the mechanism of the polymerization processes of liquid sulfur across the $λ$-transition as well as its reverse process of formation of the rings. We do so by performing ab-initio-quality molecular dynamics simulations thanks to a combination of machine learning potentials and state-of-the-art enhanced sampling techniques. With our approach, we obtain structural results that are in good agreement with the experiments and we report precious dynamical insights into the mechanisms involved in the process.

cond-mat.soft↗

arXiv2023

Multimap targeted free energy estimation

We present a new method to compute free energies at a quantum mechanical (QM) level of theory from molecular simulations using cheap reference potential energy functions, such as force fields. To overcome the poor overlap between the reference and target distributions, we generalize targeted free energy perturbation (TFEP) to employ multiple configuration maps. While TFEP maps have been obtained before from an expensive training of a normalizing flow neural network (NN), our multimap estimator allows us to use the same set of QM calculations to both optimize the maps and estimate the free energy, thus removing almost completely the overhead due to training. A multimap extension of the multistate Bennett acceptance ratio estimator is also derived for cases where samples from two or more states are available. Furthermore, we propose a one-epoch learning policy that can be used to efficiently avoid overfitting when computing the loss function is expensive compared to generating data. Finally, we show how our multimap approach can be combined with enhanced sampling strategies to overcome the pervasive problem of poor convergence due to slow degrees of freedom. We test our method on the HiPen dataset of drug-like molecules and fragments, and we show that it can accelerate the calculation of the free energy difference of switching from a force field to a DFTB3 potential by about 3 orders of magnitude compared to standard FEP and by a factor of about 8 compared to previously published nonequilibrium calculations.

physics.comp-ph↗

arXiv2023

Data Driven Classification of Ligand Unbinding Pathways

Studying the pathways of ligand-receptor binding is essential to understand the mechanism of target recognition by small molecules. The binding free energy and kinetics of protein-ligand complexes can be computed using molecular dynamics (MD) simulations, often in quantitative agreement with experiments. However, only a qualitative picture of the ligand binding/unbinding paths can be obtained through a conventional analysis of the MD trajectories. Besides, the higher degree of manual effort involved in analyzing pathways limits its applicability in large-scale drug discovery. Here we address this limitation by introducing an automated approach for analyzing molecular transition paths with a particular focus on protein-ligand dissociation. Our method is based on the dynamic time-warping (DTW) algorithm, originally designed for speech recognition. We accurately classified molecular trajectories using a very generic descriptor set of contacts or distances. Our approach outperforms manual classification by distinguishing between parallel dissociation channels, within the pathways identified by visual inspection. Most notably, we could compute exit-path-specific ligand-dissociation kinetics. The unbinding timescale along the fastest path agrees with the experimental residence time, providing a physical interpretation to our entirely data-driven protocol. In combination with appropriate enhanced sampling algorithms, this technique can be used for the initial exploration of ligand-dissociation pathways as well as for calculating path-specific thermodynamic and kinetic properties.

physics.chem-ph↗

arXiv2023

Probing the unfolded configurations of a $β$-hairpin using sketch-map

This work examines the conformational ensemble involved in $β$-hairpin folding by means of advanced molecular dynamics simulations and dimensionality reduction. A fully atomistic description of the protein and the surrounding solvent molecules is used and this complex energy landscape is sampled by means of parallel tempering metadynamics simulations. The ensemble of configurations explored is analysed using the recently proposed sketch-map algorithm. Further simulations allow us to probe how mutations affect the structures adopted by this protein. We find that many of the configurations adopted by a mutant are the same as those adopted by the wild type protein. Furthermore, certain mutations destabilize secondary structure containing configurations by preventing the formation of hydrogen bonds or by promoting the formation of new intramolecular contacts. Our analysis demonstrates that machine-learning techniques can be used to study the energy landscapes of complex molecules and that the visualizations that are generated in this way provide a natural basis for examining how the stabilities of particular configurations of the molecule are affected by factors such as temperature or structural mutations.

physics.chem-ph↗

arXiv2023

A unified framework for machine learning collective variables for enhanced sampling simulations: $\texttt{mlcolvar}$

Identifying a reduced set of collective variables is critical for understanding atomistic simulations and accelerating them through enhanced sampling techniques. Recently, several methods have been proposed to learn these variables directly from atomistic data. Depending on the type of data available, the learning process can be framed as dimensionality reduction, classification of metastable states or identification of slow modes. Here we present $\texttt{mlcolvar}$, a Python library that simplifies the construction of these variables and their use in the context of enhanced sampling through a contributed interface to the PLUMED software. The library is organized modularly to facilitate the extension and cross-contamination of these methodologies. In this spirit, we developed a general multi-task learning framework in which multiple objective functions and data from different simulations can be combined to improve the collective variables. The library's versatility is demonstrated through simple examples that are prototypical of realistic scenarios.

physics.comp-ph↗

arXiv2023

Deep Learning Collective Variables from Transition Path Ensemble

The study of the rare transitions that take place between long lived metastable states is a major challenge in molecular dynamics simulations. Many of the methods suggested to address this problem rely on the identification of the slow modes of the system which are referred to as collective variables. Recently machine learning methods have been used to learn the collective variables as functions of a large number of physical descriptors. Among many such methods Deep Targeted Discriminant Analysis has proven to be useful. This collective variable is built from data harvested in short unbiased simulation in the two basins. Here we enrich the set of data on which the Deep Targeted Discriminant Analysis collective variable is built by adding data coming from the transition path ensemble. These are collected from a number of reactive trajectories obtained using the On-the-fly Probability Enhanced Sampling Flooding method. The collective variables thus trained, lead to a more accurate sampling and faster convergence. The performance of these new collective variables is tested on a number of representative examples.

physics.chem-ph↗

arXiv2022

Water regulates the residence time of Benzamidine in Trypsin

The process of ligand-protein unbinding is crucial in biophysics. Water is an essential part of any biological system and yet, many aspects of its role remain elusive. Here, we simulate with state-of-the-art enhanced sampling techniques the binding of Benzamidine to Trypsin which is a much studied and paradigmatic ligand-protein system. We use machine learning methods to determine efficient collective coordinates for the complex non-local network of water. These coordinates are used to perform On-the-fly Probability Enhanced Sampling simulations, which we adapt to calculate also the ligand residence time. Our results, both static and dynamic, are in good agreement with experiments. We find that the presence of a water molecule located at the bottom of the binding pocket allows via a network of hydrogen bonds the ligand to be released into the solution. On a finer scale, even when unbinding is allowed, another water molecule further modulates the exit time.

physics.bio-ph↗

arXiv2022

Rare Event Kinetics from Adaptive Bias Enhanced Sampling

We introduce a novel enhanced sampling approach named OPES flooding for calculating the kinetics of rare events from atomistic molecular dynamics simulation. This method is derived from the On-the-fly-Probability-Enhanced-Sampling (OPES) approach [Invernizzi and Parrinello, JPC Lett. 2020], which has been recently developed for calculating converged free energy surfaces for complex systems. In this paper, we describe the theoretical details of the OPES flooding technique and demonstrate the application on three systems of increasing complexity: barrier crossing in a two-dimensional double well potential, conformational transition in the alanine dipeptide in gas phase, and the folding and unfolding of the chignolin polypeptide in aqueous environment. From extensive tests, we show that the calculation of accurate kinetics not only requires the transition state to be bias-free, but the amount of bias deposited should also not exceed the effective barrier height measured along the chosen collective variables. In this vein, the possibility of computing rates from biasing suboptimal order parameters has also been explored. Furthermore, we describe the choice of optimum parameter combinations for obtaining accurate results from limited computational effort.

physics.chem-ph↗

arXiv2022

Kinetics of drug release from clay using enhanced sampling methods

A key step in the development of a new drug is the design of drug-excipient complexes that lead to an optimal drug release kinetics. Computational chemistry, specifically enhanced sampling molecular dynamics methods, can play a key role in this context by minimizing the need for expensive experiments and reducing cost and time. Here, we show that recent advances in enhanced sampling methodologies can be brought to fruition in this area. We demonstrate the potential of these methodologies by simulating the release kinetics of the complex praziquantel-montmorillonite in water. Praziquantel finds promising applications in the treatment of schistosomiasis, but its biopharmaceutical profile needs to be improved and a cheap material like the montmorillonite clay would be a very convenient excipient. We simulate the drug release both from surface and interlayer space and find that the time it takes to the drug to diffuse from the interlayer space to the solution controls the release time.

physics.chem-ph↗

arXiv2022

Exploration vs Convergence Speed in Adaptive-bias Enhanced Sampling

In adaptive-bias enhanced sampling methods, a bias potential is added to the system to drive transitions between metastable states. The bias potential is a function of a few collective variables and is gradually modified according to the underlying free energy surface. We show that when the collective variables are suboptimal, there is an exploration-convergence tradeoff, and one must choose between a quickly converging bias that will lead to fewer transitions, or a slower to converge bias that can explore the phase space more efficiently but might require a much longer time to produce an accurate free energy estimate. The recently proposed On-the-fly Probability Enhanced Sampling (OPES) method focuses on fast convergence, but there are cases where fast exploration is preferred instead. For this reason, we introduce a new variant of the OPES method that focuses on quickly escaping metastable states, at the expense of convergence speed. We illustrate the benefits of this approach on prototypical systems and show that it outperforms the popular metadynamics method.

physics.comp-ph↗

arXiv2022

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 is 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.

physics.comp-ph↗

arXiv2022

Solubility of organic salts in solvent-antisolvent mixtures: A combined experimental and molecular dynamics simulations approach

We combine molecular dynamics simulations with experiments to estimate solubilities of organic salts in complex growth environments. We predict the solubility by simulations of the growth and dissolution of ions at the crystal surface kink sites at different solution concentrations. Thereby, the solubility is identified as the solution's salt concentration, where the energy of the ion pair dissolved in solution equals the energy of the ion pair crystallized at the kink sites. The simulation methodology is demonstrated for the case of anhydrous sodium acetate crystallized from various solvent-antisolvent mixtures. To validate the predicted solubilities, we have measured the solubilities of sodium acetate in-house, using an experimental setup and measurement protocol that guarantees moisture-free conditions, which is key for a hygroscopic compound like sodium acetate. We observe excellent agreement between the experimental and the computationally evaluated solubilities for sodium acetate in different solvent-antisolvent mixtures. Given the agreement and the rich data the simulations produce, we can use them to complement experimental tasks which in turn will reduce time and capital in the design of complicated industrial crystallization processes of organic salts.

cond-mat.soft↗

arXiv2021

Prediction of a supersolid phase in high-pressure deuterium

Supersolid is a mysterious and puzzling state of matter whose possible existence has stirred a vigorous debate among physicists for over 60 years. Its elusive nature stems from the coexistence of two seemingly contradicting properties, long-range order and superfluidity. We report computational evidence of a supersolid phase of deuterium under high pressure ($p >800$ GPa) and low temperature (T $<$ 1.0 K). In our simulations, that are based on bosonic path integral molecular dynamics, we observe a highly concerted exchange of atoms while the system preserves its crystalline order. The exchange processes are favoured by the soft core interactions between deuterium atoms that form a densely packed metallic solid. At the zero temperature limit, Bose-Einstein condensation is observed as the permutation probability of $N$ deuterium atoms approaches $1/N$ with a finite superfluid fraction. Our study provides concrete evidence for the existence of a supersolid phase in high-pressure deuterium and could provide insights on the future investigation of supersolid phases in real materials.

cond-mat.mtrl-sci↗

arXiv2021

Computing rates and understanding unbinding mechanism in host-guest systems

The long timescale associated with ligand residence times renders their computation challenging. Therefore, the influence of factors like solvation and steric hindrance on residence times are not fully understood. Here, we demonstrate in a set of model host-guest systems that the recently developed Gaussian Mixture Based Enhanced Sampling allows residence times to be computed and enables understanding their unbinding mechanism. We observe that guest unbinding often proceeds via a series of intermediate states that can be labelled by the number of water molecules present in the binding cavity. And in several cases the residence time is correlated to the water trapping times in the cavity.

physics.bio-ph↗

arXiv2021

A modified nudged elastic band algorithm with adaptive spring lengths

We present a modified version of the nudged elastic band (NEB) algorithm to find minimum energy paths con-necting two known configurations. We show that replacing the harmonic band-energy term with a discretized version of the Onsager-Machlup action leads to a NEB algorithm with adaptive spring lengths that automatically increase the resolution of the minimum energy path around the saddle point of the potential energy surface. The method has the same computational cost per optimization step of the standard NEB algorithm and does not introduce additional parameters. We present applications to the isomerization of alanine dipeptide, the elimination of hydrogen from ethane and the healing of a 5-77-5 defect in graphene.

physics.comp-ph↗

arXiv2021

Targeted free energy perturbation revisited: Accurate free energies from mapped reference potentials

We present an approach that extends the theory of targeted free energy perturbation (TFEP) to calculate free energy differences and free energy surfaces at an accurate quantum mechanical level of theory from a cheaper reference potential. The convergence is accelerated by a mapping function that increases the overlap between the target and the reference distributions. Building on recent work, we show that this map can be learned with a normalizing flow neural network, without requiring simulations with the expensive target potential but only a small number of single-point calculations, and, crucially, avoiding the systematic error that was found previously. We validate the method by numerically evaluating the free energy difference in a system with a double-well potential and by describing the free energy landscape of a simple chemical reaction in the gas phase.

physics.comp-ph↗