A unified framework for machine learning collective variables for enhanced sampling simulations: mlcolvar

The Journal of Chemical Physics, Journal Year: 2023, Volume and Issue: 159(1)

Published: July 6, 2023

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 directly from data. Depending on the type data available, learning process can be framed as dimensionality reduction, classification metastable states, or identification slow modes. Here, we present mlcolvar, Python library that simplifies construction their use in context contributed interface PLUMED software. The organized modularly facilitate extension cross-contamination methodologies. In this spirit, developed general multi-task framework which multiple objective functions different combined improve variables. library's versatility demonstrated simple examples are prototypical realistic scenarios.

Language: Английский

---

Enhanced Sampling of Crystal Nucleation with Graph Representation Learnt Variables

The Journal of Physical Chemistry B, Journal Year: 2024, Volume and Issue: 128(12), P. 3037 - 3045

Published: March 19, 2024

In this study, we present a graph neural network (GNN)-based learning approach using an autoencoder setup to derive low-dimensional variables from features observed in experimental crystal structures. These are then biased enhanced sampling observe state-to-state transitions and reliable thermodynamic weights. our approach, used simple convolution pooling methods. To verify the effectiveness of protocol, examined nucleation various allotropes polymorphs iron glycine their molten states. Our latent variables, when well-tempered metadynamics, consistently show between states achieve accurate rankings agreement with experiments, both which indicators dependable sampling. This underscores strength promise GNN for improved The protocol shown here should be applicable other systems

Language: Английский

---

Molecular simulation approaches to study crystal nucleation from solutions: Theoretical considerations and computational challenges

Wiley Interdisciplinary Reviews Computational Molecular Science, Journal Year: 2023, Volume and Issue: 14(1)

Published: Nov. 1, 2023

Abstract Nucleation is the initial step in formation of crystalline materials from solutions. Various factors, such as environmental conditions, composition, and external fields, can influence its outcomes rates. Indeed, controlling this rate‐determining toward phase separation critical, it significantly impact resulting material's structure properties. Atomistic simulations be exploited to gain insight into nucleation mechanisms—an aspect difficult ascertain experiments—and estimate However, microscopic nature behavior nucleating solutions when compared macroscale counterparts. An additional challenge arises inadequate timescales accessible standard molecular simulate directly; due inherent rareness events, which may apparent silico at even high supersaturations. In recent decades, simulation methods have emerged circumvent length‐ timescale limitations. not always clear method most suitable study crystal solution. This review surveys advances field, shedding light on typical mechanisms appropriateness various techniques for their study. Our goal provide a deeper understanding complexities associated with modeling solution identify areas further research. targets researchers across scientific domains, including science, chemistry, physics engineering, aims foster collaborative efforts develop new strategies understand control nucleation. article categorized under: Molecular Statistical Mechanics > Dynamics Monte‐Carlo Methods Free Energy Theoretical Physical Chemistry

Language: Английский

---

Toward On-Demand Polymorphic Transitions of Organic Crystals via Side Chain and Lattice Dynamics Engineering

Journal of the American Chemical Society, Journal Year: 2024, Volume and Issue: 146(46), P. 31911 - 31919

Published: Nov. 8, 2024

Controlling polymorphism, namely, the occurrence of multiple crystal forms for a given compound, is still an open technological challenge that needs to be addressed reliable manufacturing crystalline functional materials. Here, we devised series 13 organic crystals engineered embody molecular fragments undergoing specific nanoscale motion anticipated drive cooperative order–disorder phase transitions. By combining polarized optical microscopy coupled with heating/cooling stage, differential scanning calorimetry, X-ray diffraction, low-frequency Raman spectroscopy, and calculations (density theory dynamics), proved transitions in all systems, demonstrated how both structure lattice dynamics play crucial roles these peculiar solid-to-solid transformations. These results introduce efficient strategy design polymorphic materials endowed molecular-scale macroscopic dynamics.

Language: Английский

---

Enhanced Sampling Simulations of RNA–Peptide Binding Using Deep Learning Collective Variables

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 7, 2025

Enhanced sampling (ES) simulations of biomolecular recognition, such as binding small molecules to proteins and nucleic acid targets, protein-protein association, protein-nucleic interactions, have gained significant attention in the simulation community because their ability sample long-time scale processes. However, a key challenge implementing collective variable (CV)-based enhanced methods is selection appropriate CVs that can distinguish system's metastable states and, when biased, effectively these states. This particularly acute flexible molecule conformationally rich host simulated, peptide an RNA. In cases, large number are required capture conformations both guest well process. Using descriptors impractical any method. our work, we used recently developed deep targeted discriminant analysis (Deep-TDA) method design study cyclic peptide, L22, TAR RNA HIV, which prototypical system. The Deep-TDA CV, obtained from nonlinear combination important contact pairs between L22 backbone atoms, along with apical loop RMSD second CV were on-the-fly probability-based (OPES) reversible unbinding target. OPES delineated mechanism enabled calculation underlying free energy landscape. Our results demonstrate potential for designing complex recognition

Language: Английский

---

Long-Range Electrostatics in Serine Proteases: Machine Learning-Driven Reaction Sampling Yields Insights for Enzyme Design

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 10, 2025

Computational enzyme design is a promising technique for producing novel enzymes industrial and clinical needs. A key challenge that this faces to consistently achieve the desired activity. Fundamental studies of natural revealed critical contributions from second-shell - even more distant residues their remarkable efficiency. In particular, such organize internal electrostatic field promote reaction. Engineering fields computationally proved be strategy, which, however, has some limitations. Charged necessarily form specific patterns local interactions may exploited structural integrity. As result, it impossible probe alone by substituting amino acids. We hypothesize an approach isolates influences residues' charges other could yield deeper insights. use molecular modeling with AI-enhanced QM/MM reaction sampling implement apply model serine protease subtilisin. find negative charge 8 Å away catalytic site crucial achieving enzyme's efficiency, contributing than 2 kcal/mol lowering barrier. contrast, positive second-closest charged residue opposes efficiency raising barrier 0.8 kcal/mol. This result invites discussion into role trade-offs might have taken place in evolution enzymes. Our transferable can help investigate preorganization believe study engineering direction advance both fundamental applied enzymology lead new powerful biocatalysts.

Language: Английский

---

Theoretical and computational approaches to study crystal nucleation from solution

Published: April 25, 2023

Nucleation is the initial step towards formation of crystalline materials from solutions. Various factors, such as environmental conditions, additives, and external forces, can influence its outcomes rates. Indeed, controlling this rate-determining phase separation affect material structure properties, it crucial in a range scientific fields. In regard, atomistic simulation methods be exploited to gain insight into nucleation mechanisms - an aspect difficult ascertain experiments estimate However, microscopic nature simulations affects behaviour nucleating solutions when compared macroscopic systems. Additionally, challenge modelling solution associated with inadequacy standard molecular access timescales necessary observe crystal due inherent rareness these events. recent decades, have emerged circumvent length- timescale limitations. which method most suitable for studying not always obvious. This review summarises advances field, providing overview typical suitability different study them. By doing so, we aim provide deeper understanding complexities identify areas further research. Our targets researchers across various fields, including science, chemistry, physics engineering, will hopefully contribute developing new strategies nucleation.

Language: Английский

---

Unravelling polymorphism-driven luminescence in GFP chromophore analogues: insights into the phase transition and morphology-dependent optical waveguide properties

Journal of Materials Chemistry C, Journal Year: 2024, Volume and Issue: 12(23), P. 8368 - 8379

Published: Jan. 1, 2024

Polymorphs of GFPc analogs A and B display differences in their optical waveguiding properties 1D 2D depending on the crystal shapes. Furthermore, Form B1 demonstrates efficient capabilities even when is bent.

Language: Английский

---

Graph Neural Network-State Predictive Information Bottleneck (GNN-SPIB) approach for learning molecular thermodynamics and kinetics

Digital Discovery, Journal Year: 2024, Volume and Issue: 4(1), P. 211 - 221

Published: Nov. 28, 2024

We present a graph-based differentiable representation learning method from atomic coordinates for enhanced sampling methods to learn both thermodynamic and kinetic properties of system.

Language: Английский

---

Finite temperature string by K-means clustering sampling with order parameters as collective variables for molecular crystals: application to polymorphic transformation between β-CL-20 and ε-CL-20

Physical Chemistry Chemical Physics, Journal Year: 2023, Volume and Issue: 26(4), P. 3500 - 3515

Published: Dec. 22, 2023

Polymorphic transformation of molecular crystals is a fundamental phase transition process, and it important practically in the chemical, material, biopharmaceutical, energy storage industries.

Language: Английский

---