Machine Learning-Guided Protein Engineering

ACS Catalysis, Journal Year: 2023, Volume and Issue: 13(21), P. 13863 - 13895

Published: Oct. 13, 2023

Recent progress in engineering highly promising biocatalysts has increasingly involved machine learning methods. These methods leverage existing experimental and simulation data to aid the discovery annotation of enzymes, as well suggesting beneficial mutations for improving known targets. The field protein is gathering steam, driven by recent success stories notable other areas. It already encompasses ambitious tasks such understanding predicting structure function, catalytic efficiency, enantioselectivity, dynamics, stability, solubility, aggregation, more. Nonetheless, still evolving, with many challenges overcome questions address. In this Perspective, we provide an overview ongoing trends domain, highlight case studies, examine current limitations learning-based We emphasize crucial importance thorough validation emerging models before their use rational design. present our opinions on fundamental problems outline potential directions future research.

Language: Английский

---

Complementary probes for the electrochemical interface

Nature Reviews Chemistry, Journal Year: 2024, Volume and Issue: 8(3), P. 159 - 178

Published: Feb. 22, 2024

Language: Английский

---

DFT-Quality Adsorption Simulations in Metal–Organic Frameworks Enabled by Machine Learning Potentials

Journal of Chemical Theory and Computation, Journal Year: 2023, Volume and Issue: 19(18), P. 6313 - 6325

Published: Aug. 29, 2023

Nanoporous materials such as metal-organic frameworks (MOFs) have been extensively studied for their potential adsorption and separation applications. In this respect, grand canonical Monte Carlo (GCMC) simulations become a well-established tool computational screenings of the properties large sets MOFs. However, reliance on empirical force field potentials has limited accuracy with which can be applied to MOFs challenging chemical environments open-metal sites. On other hand, density-functional theory (DFT) is too computationally demanding routinely employed in GCMC due excessive number required function evaluations. Therefore, we propose paper protocol training machine learning (MLPs) set DFT intermolecular interaction energies (and forces) CO2 ZIF-8 site containing Mg-MOF-74, use MLPs derive isotherms from first principles. We make equivariant NequIP model demonstrated excellent data efficiency, an error below 0.2 kJ mol-1 per adsorbate was attained. Its results highly accurate heats adsorption. For dependence obtained used dispersion correction observed, where PBE-MBD performs best. Lastly, test transferability MLP trained ZIF-8, it ZIF-3, ZIF-4, ZIF-6, resulted deviations predicted Only when explicitly all ZIFs, were obtained. As proposed methodology widely applicable guest nanoporous materials, opens up possibility general-purpose perform investigations

Language: Английский

---

Operando Modeling of Zeolite-Catalyzed Reactions Using First-Principles Molecular Dynamics Simulations

ACS Catalysis, Journal Year: 2023, Volume and Issue: 13(17), P. 11455 - 11493

Published: Aug. 15, 2023

Within this Perspective, we critically reflect on the role of first-principles molecular dynamics (MD) simulations in unraveling catalytic function within zeolites under operating conditions. First-principles MD refer to methods where nuclei is followed time by integrating Newtonian equations motion a potential energy surface that determined solving quantum-mechanical many-body problem for electrons. Catalytic solids used industrial applications show an intriguing high degree complexity, with phenomena taking place at broad range length and scales. Additionally, state catalyst depend conditions, such as temperature, moisture, presence water, etc. Herein means series exemplary cases how are instrumental unravel complexity scale. Examples nature reactive species higher temperatures may drastically change compared lower active sites dynamically upon exposure water. To simulate rare events, need be combination enhanced sampling techniques efficiently sample low-probability regions phase space. Using these techniques, it shown competitive pathways conditions can discovered transition explored. Interestingly, also study hindered diffusion The clearly illustrate reveal insights into which could not using static or local approaches only few points considered (PES). Despite advantages, some major hurdles still exist fully integrate standard computational workflow use output input multiple length/time scale aim bridge reactor First all, needed allow us evaluate interatomic forces accuracy, albeit much cost currently density functional theory (DFT) methods. DFT limits attainable scales hundreds picoseconds nanometers, smaller than realistic particle dimensions encountered catalysis process. One solution construct machine learning potentials (MLPs), numerical derived from underlying data, subsequent simulations. As such, longer reached; however, quite research necessary MLPs complex systems industrially catalysts. Second, most make collective variables (CVs), mostly based chemical intuition. explore networks simulations, automatic discovery CVs do rely priori definition CVs. Recently, various data-driven have been proposed, explored systems. Lastly, investigate events. We hope rise more efficient describe PES, will future able processes catalysis. This might lead consistent dynamic description all steps─diffusion, adsorption, reaction─as they take level.

Language: Английский

---

The known-unknowns of anomalous underscreening in concentrated electrolytes

Chemical Physics Letters, Journal Year: 2024, Volume and Issue: 843, P. 141190 - 141190

Published: March 12, 2024

Electrolytes are central to life and technology but lack complete understanding. Recent experiments with highly concentrated electrolytes have revealed electrostatic decay lengths orders of magnitude larger than those predicted by theory simulation. This phenomenon, dubbed 'anomalous underscreening' its origin is still a comprehensive Herein we provide perspective over recent developments in this field discuss phenomena that, while potentially pertinent electrolyte underscreening, yet be fully explored - i.e. the 'known-unknowns' underscreening electrolytes.

Language: Английский

---

A Vision for the Future of Multiscale Modeling

ACS Physical Chemistry Au, Journal Year: 2024, Volume and Issue: 4(3), P. 202 - 225

Published: March 4, 2024

The rise of modern computer science enabled physical chemistry to make enormous progresses in understanding and harnessing natural artificial phenomena. Nevertheless, despite the advances achieved over past decades, computational resources are still insufficient thoroughly simulate extended systems from first principles. Indeed, countless biological, catalytic photophysical processes require ab initio treatments be properly described, but breadth length time scales involved makes it practically unfeasible. A way address these issues is couple theories algorithms working at different by dividing system into domains treated levels approximation, ranging quantum mechanics classical molecular dynamics, even including continuum electrodynamics. This approach known as multiscale modeling its use 60 years has led remarkable results. Considering rapid theory, algorithm design, computing power, we believe will massively grow a dominant research methodology forthcoming years. Hereby describe main approaches developed within realm, highlighting their achievements current drawbacks, eventually proposing plausible direction for future developments considering also emergence new techniques such machine learning computing. We then discuss how advanced methods could exploited critical scientific challenges, focusing on simulation complex light-harvesting processes, photosynthesis. While doing so, suggest cutting-edge paradigm consisting performing simultaneous calculations allowing various domains, with appropriate accuracy, move extend while they interact each other. Although this vision very ambitious, quick development lead both massive improvements widespread techniques, resulting and, eventually, our society.

Language: Английский

---

Understanding Defects in Amorphous Silicon with Million‐Atom Simulations and Machine Learning

Angewandte Chemie International Edition, Journal Year: 2024, Volume and Issue: 63(22)

Published: March 22, 2024

The structure of amorphous silicon (a-Si) is widely thought as a fourfold-connected random network, and yet it defective atoms, with fewer or more than four bonds, that make particularly interesting. Despite many attempts to explain such "dangling-bond" "floating-bond" defects, respectively, unified understanding still missing. Here, we use advanced computational chemistry methods reveal the complex structural energetic landscape defects in a-Si. We study an ultra-large-scale, quantum-accurate model containing million thousands individual allowing reliable defect-related statistics be obtained. combine descriptors machine-learned atomic energies develop classification different types results suggest revision established floating-bond by showing fivefold-bonded atoms a-Si exhibit wide range local environments-analogous fivefold centers coordination chemistry. Furthermore, shown (but not threefold) tend cluster together. Our provides new insights into one most studied solids, has general implications for disordered materials beyond alone.

Language: Английский

---

Ab initio guided atomistic modelling of nanomaterials on exascale high-performance computing platforms

Nano Futures, Journal Year: 2024, Volume and Issue: 8(1), P. 012501 - 012501

Published: March 1, 2024

Abstract The continuous development of increasingly powerful supercomputers makes theory-guided discoveries in materials and molecular sciences more achievable than ever before. On this ground, the incoming arrival exascale (running over 10 18 floating point operations per second) is a key milestone that will tremendously increase capabilities high-performance computing (HPC). deployment these massive platforms enable improvements accuracy scalability ab initio codes for simulation. Moreover, recent progress advanced experimental synthesis characterisation methods with atomic precision has led -based modelling to convergence terms system sizes. This it possible mimic full-scale systems silico almost without requirement inputs. article provides perspective on how computational science be further empowered by HPC, going alongside mini-review state-of-the-art HPC-aided research. Possible challenges related efficient use larger heterogeneous are commented on, highlighting importance co-design cycle. Also, some illustrative examples target applications, which could investigated detail coming years based rational nanoscale design bottom-up fashion, summarised.

Language: Английский

---

Towards understanding and controlling ultrafast dynamics in molecular photomagnets

Coordination Chemistry Reviews, Journal Year: 2023, Volume and Issue: 494, P. 215346 - 215346

Published: July 27, 2023

Manipulating magnetic materials is the cornerstone of hard drive technology. This modern information technology era has led to an explosive increase in rate data generation and storage creating urgent need achieve a new faster more efficient devices. The development ultrafast femtosecond lasers created possibility control properties using ultrashort pulses light therefore study magnetisation dynamics become one most active fields magnetism driven by both fundamental technological interest. However, major challenge this field understanding microscopic mechanisms responsible. Indeed, excited state initiated upon interaction with laser are characterised strong coupling between electronic, vibrational spin degrees freedom. especially pertinent for case single-molecule magnets, which focus review, due high density electronically states dense manifold within energy range In contribution, we discuss recent experimental theoretical developments seeking understand associated molecular photomagnets explore opportunities they offer as well outlining some future required field.

Language: Английский

---

There and back again: bridging meso- and nanoscales to understand lipid vesicle patterning

Soft Matter, Journal Year: 2024, Volume and Issue: 20(25), P. 4998 - 5013

Published: Jan. 1, 2024

We describe a complete methodology to bridge the scales between nanoscale molecular dynamics and (micrometer) mesoscale Monte Carlo simulations in lipid membranes vesicles undergoing phase separation.

Language: Английский

---